Diversity load parallel simulation method and simulation device

Abstract

The invention relates to the technical field of power system load simulation methods, in particular to a diversified load parallel simulation method and device, and the method comprises the steps: extracting load power data through historical load data, forming a fitting function based on the fitting of the load power data, and calculating the fitting load power through the fitting function; establishing a load generation function by fitting the load power; establishing a load cluster, and dividing the load cluster into a plurality of load units; and calculating the power of the plurality of load units in a multi-thread concurrent mode based on the load power generation function to obtain load simulation power, and performing load simulation power interaction through the memory bank. According to the method, through load cluster division, one thread is used for processing one load unit, so that the calculation efficiency is improved, second-level even millisecond-level load simulation power data can be generated at the same time, the change of the load is truly simulated, long-time-sequence change data needed by power grid simulation is constructed, and basic service is provided for a simulation system.

Description

Technical Field

[0001] This invention relates to the field of power system load simulation methods, and is a method and apparatus for parallel simulation of diverse loads. Background Technology

[0002] New load management is an important component of the construction of new power systems, and load management simulation is a crucial way for loads to participate in grid interaction. Simulating various load curves is a key aspect of this. Traditional load simulations are mostly based on cross-sectional data, which cannot generate long-term dynamic data. Furthermore, the number of simulated loads is often limited, the cross-sectional intervals are long, and computational efficiency is not adequately considered.

[0003] Patent application CN119182127A discloses a method, apparatus, computer device, computer-readable storage medium, and computer program product for multi-economy load forecasting of an energy system. The method includes: acquiring a dataset for multi-economy load forecasting; the dataset includes historical multi-economy load data and characteristics of influencing factors; identifying multiple energy types indicated by the historical multi-economy load data, and constructing multiple training samples for each energy type based on the dataset; training a target multi-economy load forecasting model based on the multiple training samples and weights for each energy type; and using the target multi-economy load forecasting model to output load forecasting results based on input data.

[0004] Patent application CN115879690A discloses a method and apparatus for simulating power grid load sequences. The method includes: acquiring multi-timescale characteristic parameters of the year to be simulated; calculating an initial load simulation sequence of the year to be simulated based on the monthly peak load, weekly characteristics, and daily characteristics among the multi-timescale characteristic parameters; acquiring historical load sequences and performing time-scale transformation on the historical load sequences to generate a historical load transformation sequence; and inputting the initial load simulation sequence into a pre-set load sequence simulation model to generate a target load simulation sequence of the year to be simulated. Summary of the Invention

[0005] This invention provides a method and apparatus for parallel simulation of diverse loads, which overcomes the shortcomings of the prior art. It can simulate a large number of loads, realistically simulate load changes, and improve computational efficiency.

[0006] One of the technical solutions of this invention is achieved through the following measures: a parallel simulation method for diverse loads, comprising:

[0007] Step 1: Extract load power data from historical load data, fit the load power data to form a fitting function, and use the fitting function to calculate the fitted load power;

[0008] Step 2: Establish the load generation function based on the fitted load power;

[0009] Step 3: Establish a load cluster and divide the load cluster into several load units;

[0010] Step 4: Based on the load power generation function, calculate the power of several load units in a multi-threaded concurrent manner to obtain the load simulation power, and perform load simulation power interaction through the memory library.

[0011] The following are further optimizations and / or improvements to one of the above-mentioned inventive technical solutions:

[0012] In step 1 above, the historical load power characteristic curves for each year include 26 curves, with two curves for each month: one for weekdays and one for weekends. Two additional curves are included for the Spring Festival and National Day holidays. Based on the rated capacity, load power data is extracted from the historical load power characteristic curves for each month. One power data point is extracted for each moment, and the average load power for both weekdays and non-weekdays at the same time is calculated. The average load power for both weekdays and non-weekdays at the same time is calculated using the following formula:

[0013]

[0014] In the formula, U 1i (E) represents the average power at time i on the working days of the current month; U 2i (E) represents the average power at time i on a non-working day of the month; i represents the time i on a working day or a non-working day; n represents the number of working days in the month; m represents the number of non-working days in the month; P 1i P represents the power value extracted at time i on a working day of the current month; 2i This represents the power value extracted at time i on a non-working day of the current month.

[0015] Assuming the load power characteristic curve is the air conditioning load demand response curve for a certain year, in the power data of the air conditioning load demand response curve for a certain month of a certain year, P 1i P represents the power value extracted at time i on a working day of the current month; 2i Let U represent the power value extracted at time i on a non-working day of the current month. 1i (E) is the average power at time i on a working day of a certain month of a certain year, U 2i (E) is the average power at time i on a non-working day in a certain month of a certain year.

[0016] The above fitting function was obtained as follows:

[0017] The average power at time i on the working days of the month is taken as the working day dataset, and the average power at time i on the non-working days of the month is taken as the non-working day dataset. The working day dataset and the non-working day dataset are fitted 5 times respectively to obtain the fitting function. The fitted load power of the working days and non-working days is calculated using the fitting function.

[0018] The fitting function is as follows:

[0019] P T =a5t 5 +a4t 4 +a3t 3 +a2t 2 +a1t+a0

[0020] In the formula, P T The fitted load power represents the power on weekdays or non-weekdays; t represents time; a0, a1, a2, a3, a4, and a5 are the regression coefficients obtained from higher-order fitting, respectively.

[0021] The above load power generation function is as follows:

[0022]

[0023] In the formula, P represents the simulated load power on weekdays or non-weekdays; P T This represents the fitted load power for weekdays or non-weekdays; e is the current rated power; and E is the reference coefficient.

[0024] When P T When representing the fitted load power for a workday, the simulated load power for the workday can be calculated based on the load power generation function.

[0025] When P T When representing the fitted load power on non-working days, the simulated load power on non-working days can be calculated based on the load power generation function.

[0026] The above-mentioned establishment of a load cluster, which divides the load cluster into several units, specifically includes:

[0027] The load cluster is defined as all loads within the region. The load cluster is then divided into several load units, and each load unit includes several loads.

[0028] The above calculation of the load simulation power of several load units based on the load power generation function, using a multi-threaded concurrent approach, specifically includes:

[0029] A thread is created for each load unit, and the load simulation power is calculated for the load unit corresponding to each thread in chronological order based on the load power generation function.

[0030] The second technical solution of the present invention is achieved through the following measures: a simulation device for the parallel simulation method of diverse loads described in one of the technical solutions, comprising:

[0031] The first module: extracts load power data from historical load data, forms a fitting function based on the fitting of the load power data, and calculates the fitted load power using the fitting function;

[0032] The second module: Establishing a load generation function based on the fitted load power;

[0033] The third module: Establish a load cluster and divide the load cluster into several load units;

[0034] The fourth module: Based on the load power generation function, the power of several load units is calculated in a multi-threaded concurrent manner to obtain the load simulation power, and the load simulation power is interacted with through the memory library.

[0035] The following are further optimizations and / or improvements to the second technical solution of the above invention:

[0036] The aforementioned first module includes a first unit.

[0037] Unit 1: Extract load power data from the load power characteristic curve of historical load data for each month. Extract one power data point for each moment, and calculate the average load power at the same time for both weekdays and non-weekdays in the same month. The average load power at the same time for both weekdays and non-weekdays is calculated using the following formula:

[0038]

[0039] In the formula, U 1i (E) represents the average power at time i on the working days of the current month; U 2i (E) represents the average power at time i on a non-working day of the month; i represents the time i on a working day or a non-working day; n represents the number of working days in the month; m represents the number of non-working days in the month; P 1i P represents the power value extracted at time i on a working day of the current month; 2i This represents the power value extracted at time i on a non-working day of the current month.

[0040] The first module mentioned above also includes a second unit.

[0041] Unit 2: The fitting function is obtained as follows:

[0042] The average power at time i on the working days of the month is taken as the working day dataset, and the average power at time i on the non-working days of the month is taken as the non-working day dataset. The working day dataset and the non-working day dataset are fitted 5 times respectively to obtain the fitting function. The fitted load power of the working days and non-working days is calculated using the fitting function.

[0043] The fitting function is as follows:

[0044] P T =a5t 5 +a4t 4 +a3t 3 +a2t 2 +a1t+a0

[0045] In the formula, P T The fitted load power represents the power on weekdays or non-weekdays; t represents time; a0, a1, a2, a3, a4, and a5 are the regression coefficients obtained from higher-order fitting, respectively.

[0046] The second module mentioned above includes a load simulation power calculation unit:

[0047] Load simulation power calculation unit: The load power generation function is as follows:

[0048]

[0049] In the formula, P represents the simulated load power on weekdays or non-weekdays; P T This represents the fitted load power for weekdays or non-weekdays; e is the current rated power; and E is the reference coefficient.

[0050] The third module mentioned above includes a cluster establishment unit.

[0051] Cluster Establishment Unit: Establish a load cluster by dividing the load cluster into several units, specifically including:

[0052] The load cluster is defined as all loads within the region. The load cluster is then divided into several load units, and each load unit includes several loads.

[0053] The fourth module mentioned above includes a multi-threaded computing unit.

[0054] Multi-threaded calculation unit: Based on the load power generation function, it calculates the simulated load power of several load units in a multi-threaded concurrent manner, specifically including:

[0055] A thread is created for each load unit, and the load simulation power is calculated for the load unit corresponding to each thread in chronological order based on the load power generation function.

[0056] The method described in this invention first extracts load power data from historical load characteristic curves, performs high-order fitting on the load power data to establish a fitting function, and then establishes a load power curve model. Based on this, by dividing the load clusters, one thread is used to process one load unit, thereby improving the computational efficiency and generating load simulation power data at the second or even millisecond level, which can realistically simulate load changes.

[0057] The method described in this invention can quickly generate and generate real-time load change data, thereby constructing long-term time-series change data required for power grid simulation and providing basic services for the simulation system. Attached Figure Description

[0058] Appendix Figure 1 This is a fitting curve diagram of the present invention.

[0059] Figure 1 In the diagram, blue represents weekdays and orange represents non-weekdays.

[0060] Appendix Figure 2 This is the demand response curve for a certain air conditioning load.

[0061] Appendix Figure 3 This is the load demand response curve for a certain chemical industry.

[0062] Appendix Figure 4 This is a schematic diagram of load cluster partitioning.

[0063] Appendix Figure 5 This is a schematic diagram of load concurrency in a load cluster.

[0064] Appendix Figure 6 This is a schematic diagram illustrating the process of thread computation.

[0065] Appendix Figure 7 This is a schematic diagram of load timing concurrency.

[0066] Figures 1 to 3 In the figure, the vertical axis represents power, in kW. Detailed Implementation

[0067] The present invention is not limited to the following embodiments, and the specific implementation can be determined according to the technical solution of the present invention and the actual situation.

[0068] In this invention, it should be noted that the terms "first," "second," "third," etc., are used only for the convenience of describing the invention and simplifying the description, and are not intended to indicate or imply that the modules referred to must have a specific order and operation, and therefore should not be construed as limiting the invention.

[0069] The present invention will be further described below with reference to embodiments:

[0070] Example 1: This parallel simulation method for diverse loads includes:

[0071] Step 1: Extract load power data from historical load data, fit the load power data to form a fitting function, and use the fitting function to calculate the fitted load power;

[0072] Step 2: Establish the load generation function based on the fitted load power;

[0073] Step 3: Establish a load cluster and divide the load cluster into several load units;

[0074] Step 4: Based on the load power generation function, calculate the power of several load units in a multi-threaded concurrent manner to obtain the load simulation power, and perform load simulation power interaction through the memory library.

[0075] Example 2: As an optimization of the above example, step 1 is specifically as follows:

[0076] The historical load power characteristic curves for each year include 26 curves, with two curves for each month: one for weekdays and one for weekends. Two additional curves are included for the Spring Festival and National Day holidays. Based on the rated capacity, load power data is extracted from the historical monthly load power characteristic curves. One power data point is extracted for each moment, and the average load power for both weekdays and non-weekdays at the same time is calculated. The average load power for both weekdays and non-weekdays at the same time is calculated using the following formula:

[0077]

[0078] In the formula, U 1i (E) represents the average power at time i on the working days of the current month; U 2i (E) represents the average power at time i on a non-working day of the month; i represents the time i on a working day or a non-working day; n represents the number of working days in the month; m represents the number of non-working days in the month; P 1i P represents the power value extracted at time i on a working day of the current month; 2i This represents the power value extracted at time i on a non-working day of the current month.

[0079] If one power data point is extracted every 15 minutes, and there are 96 power data points per day, then 1 ≤ i ≤ 96.

[0080] Using the average power values ​​of a certain load on weekdays and non-weekdays in a certain month, typical curves for weekdays and non-weekdays are plotted respectively. See Figure 1 .

[0081] Assuming the load power characteristic curve is the demand response curve of air conditioning load for a certain year, then for the demand response curve of air conditioning load for a certain month of a certain year (see...), Figure 2 In the power data of ), P 1i P represents the power value extracted at time i on a working day of the current month; 2i Let U represent the power value extracted at time i on a non-working day of the current month. 1i (E) is the average power at time i on a working day of a certain month of a certain year, U 2i (E) is the average power at time i on a non-working day in a certain month of a certain year.

[0082] The demand response curve for a certain chemical plant in a certain month of a certain year is shown below. Figure 3 .

[0083] Example 3: As an optimization of the above examples, the fitting function is obtained as follows:

[0084] The average power at time i on the working days of the month is taken as the working day dataset, and the average power at time i on the non-working days of the month is taken as the non-working day dataset. The working day dataset and the non-working day dataset are fitted 5 times respectively to obtain the fitting function. The fitted load power of the working days and non-working days is calculated using the fitting function.

[0085] The fitting function is as follows:

[0086] P T =a5t 5 +a4t 4 +a3t 3 +a2t 2 +a1t+a0

[0087] In the formula, P T The fitted load power represents the power on weekdays or non-weekdays; t represents time; a0, a1, a2, a3, a4, and a5 are the regression coefficients obtained from higher-order fitting, respectively.

[0088] Example 4: As an optimization of the above example, the load power generation function is as follows:

[0089]

[0090] In the formula, P represents the simulated load power on weekdays or non-weekdays; P T This represents the fitted load power for weekdays or non-weekdays; e is the current rated power; and E is the reference coefficient.

[0091] When P T When representing the fitted load power for a workday, the simulated load power for the workday can be calculated based on the load power generation function.

[0092] When P T When representing the fitted load power on non-working days, the simulated load power on non-working days can be calculated based on the load power generation function.

[0093] Example 5: As an optimization of the above examples, a load cluster is established, and the load cluster is divided into several units, specifically including:

[0094] A load cluster is defined as all loads within a region. Each load cluster is then divided into several load units, each containing several loads. For example, if a region has 20,000 loads, these 20,000 loads constitute one load cluster. Using 100 loads as one load unit, this load cluster can be divided into 200 load units. The load cluster division for a region is as follows: Figure 4 As shown, Figure 2 Three load clusters were shown.

[0095] Example 6: As an optimization of the above examples, based on the load power generation function, the load simulation power of several load units is calculated in a multi-threaded concurrent manner, specifically including:

[0096] A thread is created for each load unit, and the load simulation power is calculated for the load unit corresponding to each thread in chronological order based on the load power generation function.

[0097] Figure 5 This is a schematic diagram of load concurrency in a load cluster.

[0098] Figure 6 This is a schematic diagram of the thread computation process.

[0099] Figure 7 Schematic diagram of load timing concurrency.

[0100] Example 7: A simulation apparatus for applying the parallel simulation method for diverse loads described in one of the above examples, comprising:

[0101] The first module: extracts load power data from historical load data, forms a fitting function based on the fitting of the load power data, and calculates the fitted load power using the fitting function;

[0102] The second module: Establishing a load generation function based on the fitted load power;

[0103] The third module: Establish a load cluster and divide the load cluster into several load units;

[0104] The fourth module: Based on the load power generation function, the power of several load units is calculated in a multi-threaded concurrent manner to obtain the load simulation power, and the load simulation power is interacted with through the memory library.

[0105] Example 8: Taking load A as an example, the parallel simulation method for diverse loads is demonstrated.

[0106] Collect historical load power data for all working days of a certain month for load A. The power value of load A on a certain working day of that month is shown in Table 1.

[0107] One power data point is extracted every 15 minutes, resulting in 96 power data points per day.

[0108] Calculate the average power at the same time on all workdays of the month using the following formula:

[0109]

[0110] In the formula, U 1i (E) represents the average power at time i on the working days of the month; i represents the time i on a working day or non-working day, 1≤i≤96; n represents the number of working days in the month; P 1i This represents the power value extracted at time i on a working day of the current month.

[0111] The calculated U 1i (E) Plot a scatter plot and perform 5 fitting iterations to obtain the following fitting function:

[0112] P T = -5917.6t 5 +13697t 4 -11029t 3 +3662.5t 2 -420.5t +66.421

[0113] Based on the actual date, historical load power data is selected from the characteristic curve to determine e as the current rated power of 200kW and E as the reference power of 100kW.

[0114] The load power generation function is as follows:

[0115]

[0116] The simulated load power P at each time t is obtained based on the load power generation function.

[0117] As can be seen from the above, the method described in this invention can quickly generate and generate real-time load change data, thereby constructing long-term time-series change data required for power grid simulation and providing basic services for the simulation system.

[0118] The above technical features constitute various embodiments of the present invention, which have strong adaptability and implementation effect. Unnecessary technical features can be added or removed according to actual needs to meet the needs of different situations.

[0119] Table 1

[0120]

[0121]

Claims

1. A method for parallel simulation of diverse loads, characterized in that, include: Load power data is extracted from historical load data. Based on the fitting of the load power data, a fitting function is formed, and the fitted load power is calculated using the fitting function. The load generation function is established by fitting the load power; Establish a load cluster and divide the load cluster into several load units; Based on the load power generation function, the power of several load units is calculated in a multi-threaded concurrent manner to obtain the load simulation power, and the load simulation power is interacted with through the memory library.

2. The parallel simulation method for diverse loads according to claim 1, characterized in that, Load power data is extracted from historical load data. Specifically, load power data is extracted from the monthly load power characteristic curve of the historical load data. One power data point is extracted for each moment, and the average load power at the same moment on both weekdays and non-weekdays of the month is calculated. The average load power at the same moment on both weekdays and non-weekdays is calculated using the following formula: In the formula, U 1i (E) represents the average power at time i on the working days of the current month; U 2i (i) represents the average power at time i on a non-working day in the current month; i represents the time i on a working day or a non-working day; n represents the number of working days in the current month; m represents the number of non-working days in the current month; P 1i P represents the power value extracted at time i on a working day of the current month; 2i This represents the power value extracted at time i on a non-working day of the current month.

3. The parallel simulation method for diverse loads according to claim 2, characterized in that, The fitting function is obtained as follows: The average power at time i on the working days of the month is taken as the working day dataset, and the average power at time i on the non-working days of the month is taken as the non-working day dataset. The working day dataset and the non-working day dataset are fitted 5 times respectively to obtain the fitting function. The fitted load power of the working days and non-working days is calculated using the fitting function. The fitting function is as follows: P T =a5t 5 +a4t 4 +a3t 3 +a2t 2 +a1t+a0 In the formula, P T The fitted load power represents the power on weekdays or non-weekdays; t represents time; a0, a1, a2, a3, a4, and a5 are the regression coefficients obtained from higher-order fitting, respectively.

4. The parallel simulation method for diverse loads according to claim 3, characterized in that, The load power generation function is as follows: In the formula, P represents the simulated load power on weekdays or non-weekdays; P T This represents the fitted load power for weekdays or non-weekdays; e is the current rated power; and E is the reference coefficient.

5. The method for parallel simulation of diverse loads according to any one of claims 1 to 4, characterized in that, Establish a load cluster, and divide the load cluster into several units, specifically including: The load cluster is defined as all loads within the region. The load cluster is then divided into several load units, and each load unit includes several loads.

6. The parallel simulation method for diverse loads according to claim 5, characterized in that, Based on the load power generation function, the simulated load power of several load units is calculated in a multi-threaded concurrent manner, specifically including: A thread is created for each load unit, and the load simulation power is calculated for the load unit corresponding to each thread in chronological order based on the load power generation function.

7. A simulation apparatus applying the parallel simulation method for diverse loads as described in any one of claims 1 to 6, characterized in that, include: The first module: extracts load power data from historical load data, forms a fitting function based on the fitting of the load power data, and calculates the fitted load power using the fitting function; The second module: Establishing a load generation function based on the fitted load power; The third module: Establish a load cluster and divide the load cluster into several load units; The fourth module: Based on the load power generation function, the power of several load units is calculated in a multi-threaded concurrent manner to obtain the load simulation power, and the load simulation power is interacted with through the memory library.

8. The simulation device according to claim 7, characterized in that, The first module includes a first unit and a second unit. Unit 1: Extracting load power data from historical load data. Specifically, this involves extracting load power data from the monthly load power characteristic curve of the historical load data, extracting one power data point for each moment, and calculating the average load power at the same time on both weekdays and non-weekdays within the same month. The average load power at the same time on both weekdays and non-weekdays is calculated using the following formula: In the formula, U 1i (E) represents the average power at time i on the working days of the current month; U 2i (E) represents the average power at time i on a non-working day of the month; i represents the time i on a working day or a non-working day; n represents the number of working days in the month; m represents the number of non-working days in the month; P 1i P represents the power value extracted at time i on a working day of the current month; 2i This represents the power value extracted at time i on a non-working day of the current month; Unit 2: The fitting function is obtained as follows: The average power at time i on the working days of the month is taken as the working day dataset, and the average power at time i on the non-working days of the month is taken as the non-working day dataset. The working day dataset and the non-working day dataset are fitted 5 times respectively to obtain the fitting function. The fitted load power of the working days and non-working days is calculated using the fitting function. The fitting function is as follows: P T =a5t 5 +a4t 4 +a3t 3 +a2t 2 +a1t+a0 In the formula, P T The fitted load power represents the power on weekdays or non-weekdays; t represents time; a0, a1, a2, a3, a4, and a5 are the regression coefficients obtained from higher-order fitting, respectively.

9. The simulation device according to claim 7 or 8, characterized in that, The second module includes a load simulation power calculation unit: Load simulation power calculation unit: The load power generation function is as follows: In the formula, P represents the simulated load power on weekdays or non-weekdays; P T This indicates the fitted load power for weekdays or non-weekdays; e represents the current rated power. E is a reference coefficient; Or / and, the third module includes a cluster establishment unit, Cluster Establishment Unit: Establish a load cluster by dividing the load cluster into several units, specifically including: The load cluster is defined as all loads within the region. The load cluster is then divided into several load units, and each load unit includes several loads.

10. The simulation device according to claim 9, characterized in that, The fourth module includes a multi-threaded computing unit. Multi-threaded calculation unit: Based on the load power generation function, it calculates the simulated load power of several load units in a multi-threaded concurrent manner, specifically including: A thread is created for each load unit, and the load simulation power is calculated for the load unit corresponding to each thread in chronological order based on the load power generation function.

Images