Power system probabilistic load flow calculation method based on calculation force load random model

Abstract

The invention discloses a power system probabilistic load flow calculation method based on a calculation force load random model. The method comprises the following steps: acquiring illumination intensity data, wind speed data and basic data of a power grid system; establishing a probability distribution model of wind power and photoelectric output and an output probability distribution model of a common load in the line; establishing a probability model of the data center load; constructing a linearized probabilistic power flow model of the power system structure; according to the linearized probabilistic power flow model, calculating the original moment and the center distance of an input variable to obtain the semi-invariant of each order, and further obtaining a state variable and the semi-invariant of the branch power flow; and obtaining a probability density function and a cumulative distribution function of the state variable and the branch power flow according to the state variable and the semi-invariant of the branch power flow. The method not only considers the photovoltaic and wind power uncertain power injection of the power distribution network, but also considers the random model of the calculation load of the data center, so that the result is closer to reality, the calculation accuracy is high, and the speed is high.

Description

Technical Field

[0001] This invention belongs to the field of power systems and relates to power system probabilistic power flow calculation technology, specifically to a power system probabilistic power flow calculation method based on a computing load stochastic model. Background Technology

[0002] With the advancement of the "dual carbon" target, the proportion of new energy sources such as wind power and photovoltaics in the power system has increased significantly. The intermittent and random nature of their output poses new challenges to the stable operation of the power grid. Meanwhile, the power consumption of data centers, the cornerstone of the digital economy, continues to grow, and the dramatic fluctuations in their computing load further exacerbate the uncertainty of power system load. Against this backdrop, traditional deterministic power flow calculations are no longer sufficient to accurately depict the operating state of the power system. Therefore, probabilistic power flow calculation methods that can effectively handle multiple random factors have become a key research focus.

[0003] Currently, probabilistic power flow calculation methods are mainly divided into three categories: simulation methods (such as Monte Carlo simulation), approximation methods (such as point estimation methods), and analytical methods (such as semi-invariant methods). Monte Carlo simulation calculates power flow based on a large number of input samples, offering high accuracy but at a huge computational cost. Point estimation methods transform input random variables into deterministic inputs through equivalent transformations, and then estimate the power flow output. While computationally efficient, these methods have limitations in handling higher-order moments and obtaining complete probability distributions. Traditional analytical methods perform convolution calculations based on the probability distribution functions of random variables, facing challenges in computational complexity and accuracy. Existing research primarily seeks breakthroughs in model accuracy, computational efficiency, and handling variable correlations, but often fails to fully consider the unique stochastic characteristics of novel loads such as data centers. Therefore, developing an efficient probabilistic power flow calculation method that can accurately integrate stochastic models of data center computing power loads is of significant value for the planning and safe operation of new power systems. Summary of the Invention

[0004] Purpose of the invention: To overcome the shortcomings of existing technologies, this invention provides a probabilistic power flow calculation method for power systems based on a stochastic model of computing load. This method not only considers the uncertainties of photovoltaic and wind power in the power grid, but also the random fluctuations of computing load in data centers. It can effectively calculate the power flow probability density function and cumulative distribution function of the power grid, and the calculation results are highly accurate and fast, which helps to conduct safety and stability analysis and preventive control of the power system.

[0005] Technical Solution: To achieve the above objectives, this invention provides a probabilistic power flow calculation method for power systems based on a computational load stochastic model, comprising the following steps:

[0006] S1: Acquire solar intensity data, wind speed data, and basic data of the power grid system at new energy power plants;

[0007] S2: Based on the distribution types and parameters of wind and solar new energy, establish a probability distribution model for wind power and solar power output. At the same time, use the normal distribution to describe the load power distribution within the line and establish a probability distribution model for the output of ordinary loads within the line.

[0008] S3: Based on the random distribution type of basic data and computing load in data centers, establish a probabilistic model of data center load;

[0009] S4: Based on the probability distribution models of wind power and solar power output, the probability distribution models of output of ordinary loads in the line, and the probability models of data center loads, construct a linearized probabilistic power flow model of the power system structure.

[0010] S5: Based on the linearized probabilistic power flow model, the semi-invariants of each order are obtained by calculating the raw moments and center moments of the input variables, and then the semi-invariants of the state variables and branch power flow are obtained.

[0011] S6: Based on the semi-invariants of the state variables and branch power flow, obtain the probability density function and cumulative distribution function of the state variables and branch power flow.

[0012] Furthermore, the probability distribution model for wind power and solar power output in step S2 includes a distributed wind power node probability model and a distributed photovoltaic node probability model.

[0013] Furthermore, the method for establishing the distributed wind power node probability model in step S2 includes:

[0014] Using the Weibull distribution to fit the wind speed, the probability density function of the wind speed is expressed as:

[0015]

[0016] Where c and k represent the scale parameter and shape parameter of the Weibull distribution, respectively; v represents the wind speed in the wind farm.

[0017] Based on the wind speed model, the probabilistic model of the wind turbine's output power is obtained:

[0018]

[0019] in, , ; , These are the output power and rated power of the wind turbine, respectively. , , These represent the fan's cut-in, cut-out, and rated wind speed, respectively.

[0020] Furthermore, the method for establishing the distributed photovoltaic node probability model in step S2 includes:

[0021] The probability distribution of light intensity follows a Beta distribution, and the active power output of a photovoltaic power generator is expressed as:

[0022]

[0023]

[0024] Where r is the actual light intensity; This represents the maximum light intensity. The total area of ​​the square array. This represents the total photoelectric conversion efficiency of the array;

[0025] The probability density function of the active power output by a photovoltaic power plant:

[0026]

[0027] in, is the Gamma function; parameters a and b are two shape parameters of the Beta distribution, namely the first shape parameter and the second shape parameter, respectively.

[0028] Furthermore, the output probability distribution model of ordinary loads within the line in step S2 is expressed as follows:

[0029] For load uncertainties in power grid lines and load fluctuations with region and time, a normal distribution function model is used to fit the load output model:

[0030]

[0031] in, and These are the expected and standard deviation of the load output, respectively.

[0032] Furthermore, the establishment of the probabilistic model for data center load in step S3 includes:

[0033] The total power of data center i during time period t is:

[0034]

[0035] in, For IT equipment power, Let be the cooling system power of data center i during time period t. For the power of other systems in data center i;

[0036] IT equipment power :

[0037]

[0038] in, The power of a single server. Network transmission power;

[0039] Considering that typical server modeling uses a linear model, which introduces some errors, an empirical nonlinear model is proposed, namely, the power of a single server. Represented as:

[0040]

[0041] in, and These represent the idle and full load power of a single server, respectively. This represents the CPU utilization rate of the server, where These are calibration parameters that reduce and minimize the squared error that needs to be obtained experimentally; this nonlinear empirical model tracks dynamic power consumption very well. Server linear models typically have errors below 5%, while this model's error is below 1%.

[0042] Assuming all servers within the data center are homogeneous, the average CPU utilization of the servers can be expressed as:

[0043]

[0044] in, Let be the number of servers active in data center i during time period t in scenario s; Let be the average arrival rate of interactive workloads from the front-end agent in data center i during time period t; The average service rate at which servers within data center i handle workloads;

[0045] The total power of the server is expressed as:

[0046]

[0047] Based on the server power model, using a given power coefficient and The total power consumption of the entire data center can be calculated as follows:

[0048] .

[0049] Furthermore, the construction of the linearized probabilistic power flow model of the power system structure in step S4 includes:

[0050] The power flow equations for nodes and branches are linearized and expressed in polar coordinates:

[0051]

[0052] Where W represents the power of the power flow pair node, and Z represents the power flow of the branch;

[0053] Expanding by Taylor series and ignoring terms of degree two and above, we get:

[0054]

[0055]

[0056] Simplifying, we get:

[0057]

[0058] in, Here is the Jacobian matrix used in the last iteration of the Newton-Lafrian method, where ; A vector of corrections to node state variables. For the random disturbance component of the branch power flow, Represents the unbalanced power vector at the nodes;

[0059] Power flow calculations are performed using the Newton-Layer method, by calculating the expected values ​​of the node voltages operating at the reference point. Expected value of branch tidal current And Jacobi matrix Further solve the sensitivity matrix and .

[0060] Furthermore, in step S5, the semi-invariants of the injected power are first obtained, and then the improved semi-invariant method is used to obtain the semi-invariants of the state variables and the branch power flow.

[0061] When calculating the semi-invariants of injected power, the impact of not only wind and solar renewable energy and ordinary loads, but also the random loads of the data center are considered. The sum of the distribution functions of several independent random variables, calculated by convolving them, is the distribution function of their sum. Therefore, the random variables of injected power for each node are:

[0062]

[0063] in, , , and These represent the changes in random variables such as data center load, general load, wind power output, and photovoltaic power output, respectively.

[0064] Because it involves complex convolution calculations, the computation time is too long, so an improved semi-invariant method is adopted.

[0065] Based on the additive property of the semi-invariant method:

[0066]

[0067] Then by The state variables are obtained by calculating the semi-invariants of each order. The semi-invariants of each order:

[0068]

[0069] Branch Road Trend The semi-invariants of each order are obtained by the following equation:

[0070]

[0071] in, , Each is a matrix , A matrix formed by the k powers of the elements in the matrix.

[0072] Further, in step S6, the probability density function and cumulative distribution function are obtained by fitting using Gram-Charlier series expansion, specifically including:

[0073] In stochastic production simulation of power systems, the Gram-Charlier series expands the distribution function of a random variable into a series composed of the derivatives of a normally distributed random variable, and the coefficients of the series are expressed as expressions for the semi-invariants of the random variable. To simplify the form of the series, we define:

[0074]

[0075] in, For ν-order normalized semi-invariants, Let X be the standard deviation; for any random variable X, suppose the expected value and standard deviation are respectively... and Its standardized form is

[0076]

[0077] set up and Standard random variables The cumulative distribution function and probability density function, and ;

[0078] but and The Gram-Charlier series expansion is as follows:

[0079]

[0080]

[0081] in, It is the standard normal distribution density function; For Hermite polynomials; ! represents factorial;

[0082] Finally, the probability distribution of voltage and power at each node of the power system is obtained.

[0083] Beneficial Effects: Compared with existing technologies, this invention employs an improved semi-invariant probabilistic power flow calculation method, replacing convolution calculation with addition, thus simplifying the algorithm. This invention considers not only the uncertain power injection from photovoltaic and wind power in the distribution network but also the stochastic model of data center computing load, making the results closer to reality, with high calculation accuracy and speed. The method of this invention is of great significance for the security analysis and prevention control of complex power grid systems after large-scale distributed power sources and data centers are connected to the grid. Attached Figure Description

[0084] Figure 1 This is a schematic flowchart of the method of the present invention;

[0085] Figure 2 A schematic diagram of the IEEE 33-node system;

[0086] Figure 3 Comparison of voltage amplitude probability density at node 6;

[0087] Figure 4 A comparison chart of the cumulative voltage amplitude distribution at node 6. Detailed Implementation

[0088] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading this invention, any modifications of the invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.

[0089] Example 1:

[0090] like Figure 1 As shown, this embodiment provides a method for calculating the probabilistic power flow of a power system based on a stochastic model of data center computing power load, including the following steps:

[0091] S1: Acquire solar intensity data, wind speed data, and basic data of the power grid system at new energy power plants;

[0092] S2: Based on the distribution types and parameters of wind and solar new energy, establish a probability distribution model for wind power and solar power output. At the same time, use the normal distribution to describe the load power distribution within the line and establish a probability distribution model for the output of ordinary loads within the line.

[0093] The probability distribution models for wind and solar power output include distributed wind power node probability models and distributed photovoltaic node probability models.

[0094] Methods for establishing distributed wind power node probability models include:

[0095] Using the Weibull distribution to fit the wind speed, the probability density function of the wind speed is expressed as:

[0096]

[0097] Where c and k represent the scale parameter and shape parameter of the Weibull distribution, respectively; v represents the wind speed in the wind farm.

[0098] Based on the wind speed model, the probabilistic model of the wind turbine's output power is obtained:

[0099]

[0100] in, , ; , These are the output power and rated power of the wind turbine, respectively. , , These represent the fan's cut-in, cut-out, and rated wind speed, respectively.

[0101] Methods for establishing distributed photovoltaic node probabilistic models include:

[0102] The probability distribution of light intensity follows a Beta distribution, and the active power output of a photovoltaic power generator is expressed as:

[0103]

[0104]

[0105] Where r is the actual light intensity; This represents the maximum light intensity. The total area of ​​the square array. This represents the total photoelectric conversion efficiency of the array;

[0106] The probability density function of the active power output by a photovoltaic power plant:

[0107]

[0108] in, is the Gamma function; parameters a and b are two shape parameters of the Beta distribution, namely the first shape parameter and the second shape parameter, respectively.

[0109] The output probability distribution model for ordinary loads within the line is expressed as follows:

[0110] For load uncertainties in power grid lines and load fluctuations with region and time, a normal distribution function model is used to fit the load output model:

[0111]

[0112] in, and These are the expected and standard deviation of the load output, respectively.

[0113] S3: Based on the random distribution type of basic data and computing load in data centers, establish a probabilistic model of data center load;

[0114] Establish probabilistic models for data center nodes, especially when establishing power models for individual servers in data center IT equipment, and consider more accurate nonlinear models;

[0115] The total power of data center i during time period t is:

[0116]

[0117] in, For IT equipment power, Let be the cooling system power of data center i during time period t. For the power of other systems in data center i;

[0118] IT equipment power :

[0119]

[0120] in, The power of a single server. Network transmission power;

[0121] Considering that typical server modeling uses a linear model, which introduces some errors, an empirical nonlinear model is proposed, namely, the power of a single server. Represented as:

[0122]

[0123] in, and These represent the idle and full load power of a single server, respectively. This represents the CPU utilization rate of the server, where These are calibration parameters that reduce and minimize the squared error that needs to be obtained experimentally; this nonlinear empirical model tracks dynamic power consumption very well. Server linear models typically have errors below 5%, while this model's error is below 1%.

[0124] Assuming all servers within the data center are homogeneous, the average CPU utilization of the servers can be expressed as:

[0125]

[0126] in, Let be the number of servers active in data center i during time period t in scenario s; Let be the average arrival rate of interactive workloads from the front-end agent in data center i during time period t; The average service rate at which servers within data center i handle workloads;

[0127] The total power of the server is expressed as:

[0128]

[0129] A data center network consists of two parts: transmission links and transmission equipment. The dynamic range of the power of both is very small. Therefore, the network transmission power of data center i can be regarded as a fixed value.

[0130] The largest and most variable component of data center energy consumption is server energy consumption, while the energy consumption of other devices is relatively fixed compared to server energy consumption.

[0131] The total power of a data center is positively correlated with the power of its servers. This is based on a server power model, using a given power factor. and The total power consumption of the entire data center can be calculated as follows:

[0132]

[0133] S4: Based on the probability distribution models of wind power and solar power output, the probability distribution models of output of ordinary loads in the line, and the probability models of data center loads, construct a linearized probabilistic power flow model of the power system structure.

[0134] The power flow equations for nodes and branches are linearized and expressed in polar coordinates:

[0135]

[0136] in, Inject power into the node, For branch road currents;

[0137] Expanding by Taylor series and ignoring terms of degree two and above, we get:

[0138]

[0139]

[0140] Simplifying, we get:

[0141]

[0142] in, Here is the Jacobian matrix used in the last iteration of the Newton-Lafrian method, where ; A vector of corrections to node state variables. For the random disturbance component of the branch power flow, Represents the unbalanced power vector at the nodes;

[0143] Power flow calculations are performed using the Newton-Layer method, by calculating the expected values ​​of the node voltages operating at the reference point. Expected value of branch tidal current And Jacobi matrix Further solve the sensitivity matrix and .

[0144] S5: Based on the linearized probabilistic power flow model, the semi-invariants of each order are obtained by calculating the raw moments and center moments of the input variables, and then the semi-invariants of the state variables and branch power flow are obtained.

[0145] First, the semi-invariants of the injected power are obtained, and then the improved semi-invariant method is used to obtain the semi-invariants of the state variables and the branch power flow.

[0146] When calculating the semi-invariants of injected power, the impact of not only wind and solar renewable energy and ordinary loads, but also the random loads of data centers is considered; the sum of the distribution functions of several independent random variables is calculated by convolving them, and is the distribution function of their sum. Therefore, the random variables of injected power for each node are:

[0147]

[0148] in, , , and These represent the changes in random variables such as data center load, general load, wind power output, and photovoltaic power output, respectively.

[0149] Because it involves complex convolution calculations, the computation time is too long, so an improved semi-invariant method is adopted.

[0150] Based on the additive property of the semi-invariant method:

[0151]

[0152] Then by The state variables are obtained by calculating the semi-invariants of each order. The semi-invariants of each order:

[0153]

[0154] Branch Road Trend The semi-invariants of each order are obtained by the following equation:

[0155]

[0156] in, , Each is a matrix , A matrix formed by the k powers of the elements in the matrix.

[0157] S6: Based on the semi-invariants of the state variables and branch power flow, obtain the probability density function and cumulative distribution function of the state variables and branch power flow, and finally perform probability power flow calculation based on the plotted probability density curve.

[0158] In this embodiment, the probability density function and cumulative distribution function are obtained by fitting using Gram-Charlier series expansion, specifically including:

[0159] In stochastic production simulation of power systems, the Gram-Charlier series expands the distribution function of a random variable into a series composed of the derivatives of a normally distributed random variable, and the coefficients of the series are expressed as expressions for the semi-invariants of the random variable. To simplify the form of the series, we define:

[0160]

[0161] in, For ν-order normalized semi-invariants, Let X be the standard deviation; for any random variable X, suppose the expected value and standard deviation are respectively... and Its standardized form is

[0162]

[0163] set up and Standard random variables The cumulative distribution function and probability density function, and ;

[0164] but and The Gram-Charlier series expansion is as follows:

[0165]

[0166]

[0167] in, It is the standard normal distribution density function; For Hermite polynomials; ! represents factorial;

[0168] Finally, the probability distribution of voltage and power at each node of the power system is obtained.

[0169] Example 2:

[0170] To verify the effectiveness and efficacy of the method of the present invention, the following experiments and analyses were conducted in this embodiment:

[0171] like Figure 2 As shown, this embodiment is based on the IEEE 33-node system, which includes 33 nodes, namely... Figure 2 The nodes are numbered 1 to 33, with a base power of 10MVA. Assume there is a data center load at node 18, a wind turbine at node 31, and a photovoltaic generator at node 32.

[0172] The wind speed at the location of the wind farm follows a Weibul distribution with a shape parameter k of 4 and a scale parameter c of 10.35 m / s. The wind farm consists of wind turbines with a cut-in wind speed of 3 m / s, a rated wind speed of 13 m / s, and a cut-out wind speed of 25 m / s.

[0173] A 20MW photovoltaic power station is connected at node 31, with 40 photovoltaic modules having a rated power of 0.5MW each. The power output of the photovoltaic power station adopts a Beta distribution with shape factors of 0.52 and 2.6.

[0174] The load at each node follows a normal distribution, with the mean being the predicted load value and the standard deviation being 5% of the mean.

[0175] Assume the internet service company has one front-end proxy. The workload arrival rate of the front-end proxy follows a normal distribution with a mean equal to the predicted value and a standard deviation of 5% of the mean. The annual continuous curve of the workload arrival rate has the same shape as the node load. The data center is configured with servers at twice the peak workload. The average service rate of the servers handling the workload is 20 requests / s. The idle power and full load power of the active servers are 0.1kW and 0.2kW, respectively. The power factor 'a' of the data center is 1.4, and the power factor 'b' is 0.

[0176] First, based on the relevant parameters and distribution types of wind farms and photovoltaic power stations, a probabilistic model of their power output is established. At the same time, a probabilistic model of their load is established based on the computing power load of data centers.

[0177] Then, based on the structure of the power system, a probabilistic power flow model is constructed and linearized. The semi-invariants of injected power are obtained using models of wind, solar, and data centers, and the semi-invariants of state variables and branch power flow are obtained using an improved semi-invariant method.

[0178] Finally, the probability density functions and cumulative distribution functions of the state variables and branch power flows are obtained through Gram-Chariler series expansion, and compared with the traditional Monte Carlo Simulation (MCS) method. The comparison results are as follows: Figure 3 and Figure 4 As shown.

[0179] from Figure 3 It can be seen that the voltage amplitude at node 6 is mainly concentrated in the range of 0.94~0.96 pu, with a peak value around 0.95 pu. The overall distribution is unimodal and approximately symmetrical, consistent with the typical probabilistic characteristics of node voltages in power systems. The peak position, distribution range, and overall profile of the CM (Gram-Charlier) curve and the MCS histogram are highly consistent, accurately reproducing the voltage amplitude probability density distribution obtained from Monte Carlo simulation. This verifies that the method of this invention can fit the distribution of node voltage amplitudes well.

[0180] from Figure 4 It can be seen that the cumulative probability of the voltage amplitude at node 6 increases monotonically with the voltage amplitude, rising rapidly in the range of 0.94 to 0.96 pu, and eventually approaching 1, which conforms to the typical characteristics of a unimodal cumulative probability distribution. The overall trend of the CM (Gram-Charlier) curve and the MCS dashed line highly overlaps, verifying that the method of this invention can accurately reproduce the cumulative distribution law of voltage amplitude obtained by Monte Carlo simulation.

Claims

1. A method for calculating probabilistic power flow in a power system based on a stochastic model of computational load, characterized in that, Includes the following steps: S1: Acquire solar intensity data, wind speed data, and basic data of the power grid system at new energy power plants; S2: Based on the distribution types and parameters of wind and solar new energy, establish a probability distribution model for wind power and solar power output. At the same time, use the normal distribution to describe the load power distribution within the line and establish a probability distribution model for the output of ordinary loads within the line. S3: Based on the random distribution type of basic data and computing load in data centers, establish a probabilistic model of data center load; S4: Based on the probability distribution models of wind power and solar power output, the probability distribution models of output of ordinary loads in the line, and the probability models of data center loads, construct a linearized probabilistic power flow model of the power system structure. S5: Based on the linearized probabilistic power flow model, the semi-invariants of each order are obtained by calculating the raw moments and center moments of the input variables, and then the semi-invariants of the state variables and branch power flow are obtained. S6: Based on the semi-invariants of the state variables and branch power flow, obtain the probability density function and cumulative distribution function of the state variables and branch power flow.

2. The power system probabilistic power flow calculation method based on a computational load stochastic model according to claim 1, characterized in that, The probability distribution models for wind power and solar power output in step S2 include distributed wind power node probability models and distributed solar power node probability models.

3. The power system probabilistic power flow calculation method based on a computational load stochastic model according to claim 2, characterized in that, The method for establishing the distributed wind power node probability model in step S2 includes: Using the Weibull distribution to fit the wind speed, the probability density function of the wind speed is expressed as: ； Where c and k represent the scale parameter and shape parameter of the Weibull distribution, respectively; v represents the wind speed in the wind farm. Based on the wind speed model, the probabilistic model of the wind turbine's output power is obtained: ； in, , ; , These are the output power and rated power of the wind turbine, respectively. , , These represent the fan's cut-in, cut-out, and rated wind speed, respectively.

4. The power system probabilistic power flow calculation method based on a computational load stochastic model according to claim 3, characterized in that, The method for establishing the distributed photovoltaic node probability model in step S2 includes: The probability distribution of light intensity follows a Beta distribution, and the active power output of a photovoltaic power generator is expressed as: ； ； Where r is the actual light intensity; This represents the maximum light intensity. The total area of ​​the square array. This represents the total photoelectric conversion efficiency of the array; The probability density function of the active power output by a photovoltaic power plant: ； in, is the Gamma function; parameters a and b are two shape parameters of the Beta distribution, namely the first shape parameter and the second shape parameter, respectively.

5. The power system probabilistic power flow calculation method based on a computational load stochastic model according to claim 4, characterized in that, The output probability distribution model of ordinary loads within the line in step S2 is expressed as follows: For load uncertainties in power grid lines and load fluctuations with region and time, a normal distribution function model is used to fit the load output model: ； in, and These are the expected and standard deviation of the load output, respectively.

6. The power system probabilistic power flow calculation method based on a computational load stochastic model according to claim 5, characterized in that, The establishment of the probabilistic model for data center load in step S3 includes: The total power of data center i during time period t is: ； in, For IT equipment power, Let be the cooling system power of data center i during time period t. For the power of other systems in data center i; IT equipment power : ； in, The power of a single server. Network transmission power; Power of a single server Represented as: ； in, and These represent the idle and full load power of a single server, respectively. This represents the CPU utilization rate of the server, where These are calibration parameters; Assuming all servers within the data center are homogeneous, the average CPU utilization of the servers can be expressed as: ； in, Let be the number of servers active in data center i during time period t in scenario s; Let be the average arrival rate of interactive workloads from the front-end agent in data center i during time period t; The average service rate at which servers within data center i handle workloads; The total power of the server is expressed as: ； Based on the server power model, using a given power coefficient and The total power consumption of the entire data center can be calculated as follows: 。 7. The method for calculating probabilistic power flow in a power system based on a stochastic model of computational load according to claim 6, characterized in that, The construction of the linearized probabilistic power flow model of the power system structure in step S4 includes: The power flow equations for nodes and branches are linearized and expressed in polar coordinates: ； Where W represents the power of the power flow pair node, and Z represents the power flow of the branch; Expanding by Taylor series and ignoring terms of degree two and above, we get: ； ； Simplifying, we get: ； in, Here is the Jacobian matrix used in the last iteration of the Newton-Lafrian method, where ; A vector of corrections to node state variables. For the random disturbance component of the branch power flow, Represents the unbalanced power vector at the nodes; Power flow calculations are performed using the Newton-Layer method, by calculating the expected values ​​of the node voltages operating at the reference point. Expected value of branch tidal current And Jacobi matrix Further solve the sensitivity matrix and .

8. The method for calculating probabilistic power flow in a power system based on a stochastic model of computational load according to claim 7, characterized in that, In step S5, the semi-invariants of the injected power are first obtained, and then the improved semi-invariant method is used to obtain the semi-invariants of the state variables and the branch power flow. Random variable for node injected power: ； in, , , and These represent the changes in random variables such as data center load, general load, wind power output, and photovoltaic power output, respectively. According to the additive property of the semi-invariant method: ； Then by The state variables are obtained by calculating the semi-invariants of each order. The semi-invariants of each order: ； Branch Road Trend The semi-invariants of each order are obtained by the following equation: ； in, , Each is a matrix , A matrix formed by the k powers of the elements in the matrix.

9. The method for calculating probabilistic power flow in a power system based on a stochastic model of computational load according to claim 8, characterized in that, Step S6, which uses Gram-Charlier series expansion to fit and obtain the probability density function and cumulative distribution function, specifically includes: In stochastic production simulation of power systems, the Gram-Charlier series expands the distribution function of a random variable into a series composed of the derivatives of a normally distributed random variable, and the coefficients of the series are expressed as expressions for the semi-invariants of the random variable. To simplify the form of the series, we define: ； in, For ν-order normalized semi-invariants, Let X be the standard deviation; for any random variable X, suppose the expected value and standard deviation are respectively... and Its standardized form is ； set up and Standard random variables The cumulative distribution function and probability density function, and ; but and The Gram-Charlier series expansion is as follows: ； ； in, It is the standard normal distribution density function; For Hermite polynomials; ! represents factorial; Finally, the probability distribution of voltage and power at each node of the power system is obtained.

Images