Voltage out-of-limit risk assessment method for wind-solar access power distribution system

Abstract

The invention discloses a voltage out-of-limit risk assessment method for a wind and light access power distribution system, and the method comprises the following steps: 1, carrying out the time-segmented processing of the irradiance and wind speed in a distributed power supply probability model, and carrying out the day and night segmented sampling through Monte Carlo, so as to generate a large number of wind and light scenes; 2, carrying out load flow calculation on a large number of wind and light scenes generated in the step 1 to obtain voltage data of each node; 3, fitting the voltage data in the step 2 by adopting a Gaussian mixture model to obtain a voltage probability density function; and 4, according to the voltage probability density function in the step 3, further calculating the voltage out-of-limit probability and the voltage comprehensive evaluation index of the node. According to the method, no matter how complex the distribution form of voltage data is, real distribution can be approached by adjusting parameters of each Gaussian component, so that the voltage out-of-limit probability and the voltage comprehensive risk index of each node are accurately calculated.

Description

Technical Field

[0001] The present invention belongs to the field of voltage over-limit risk assessment for distribution networks, and in particular relates to a method for voltage over-limit risk assessment for wind-solar access distribution systems. Background Art

[0002] The growth of demand for new energy and the increasing public concern about environmental issues have promoted the rapid development of distributed power generation. With the country's increasing support for new energy and the continuous decline in construction costs, a large number of small, decentralized distributed power sources are rapidly connected to the distribution network. In some areas, the access ratio of distributed photovoltaic and wind power generation is already high, which has a significant impact on the safe, stable operation and economy of the local power grid, specifically manifested in problems such as grid voltage over-limit, power reverse transmission and line overload. In particular, excessive voltage not only threatens the safe and stable operation of the line, but also affects the distribution network's ability to accept power generation and the system's operating efficiency. The overvoltage phenomenon is mainly caused by the instability and intermittent characteristics of distributed energy generation, and the existing power grid operation mode is difficult to adapt to the access characteristics of distributed photovoltaic and wind power generation. When the power generation of photovoltaic and wind power generation systems exceeds the actual demand of the power grid, it may cause voltage increase, which in turn causes a series of safety hazards, such as damage to electrical equipment and waste of electricity.

[0003] Therefore, the random volatility and intermittency of wind power and photovoltaic power increase the complexity and uncertainty of the distribution system, bringing more severe challenges to the safe operation of the distribution system. Therefore, it is of great significance to comprehensively and quickly evaluate the voltage risk of the distribution system containing wind power and photovoltaic power, and its results can provide reference information for the risk assessment of new energy grid connection in the future. Summary of the invention

[0004] The purpose of the present invention is to provide a method for assessing voltage over-limit risk in a wind-solar access distribution system. No matter how complex the distribution of voltage data is, the parameters of each Gaussian component can be adjusted to approximate the actual distribution, thereby accurately calculating the voltage over-limit probability and voltage comprehensive risk index of each node.

[0005] The technical solution adopted by the present invention is a method for assessing voltage over-limit risk of wind-solar access power distribution system, comprising the following steps:

[0006] Step 1: First, the irradiance and wind speed in the distributed power probability model are processed by time period, and then Monte Carlo is used to perform day and night segment sampling to generate a large number of wind and solar scenes;

[0007] Step 2: Perform power flow calculation on a large number of wind and solar scenarios generated in step 1 to obtain voltage data of each node;

[0008] Step 3: Use the mixed Gaussian model to fit the voltage data in step 2 to obtain the voltage probability density function;

[0009] Step 4: Based on the voltage probability density function in step 3, further calculate the voltage over-limit probability and voltage comprehensive evaluation index of the node.

[0010] The present invention is also characterized in that:

[0011] Step 1 is as follows:

[0012] Step 1.1: In the probability model of PV power station output, the relationship between PV power station output and light intensity is:

[0013]

[0014] Where: γ max is the maximum value of light intensity; γ is the actual value of light intensity; Γ is the Gamma function; α and β are both shape parameters of Beta distribution;

[0015] In this probability model, the average and standard values ​​of the light intensity required for photovoltaic cells to generate power within a certain period of time are approximately:

[0016]

[0017] Where: μ is the mean; δ 2 is the variance; α and β are the shape parameters of Beta distribution;

[0018] The output power of the photovoltaic cell is:

[0019]

[0020] Where: P M is the actual output power of the photovoltaic cell; A is the total area of ​​the photovoltaic cell; η is the photoelectric conversion power of the photovoltaic cell; R M is the maximum output power of the photovoltaic cell; γ max is the maximum value of light intensity;

[0021] The probability density of photovoltaic output is obtained from equations (1) to (3):

[0022]

[0023] Step 1.2: Divide the photovoltaic beta parameters calculated at the hourly level into two segments, generating daytime and nighttime periods; the daytime period is 6:00-18:00, and the nighttime periods are 0:00-6:00 and 18:00-24:00; then, recombine the nighttime period 18:00-24:00 with the time period 0:00-6:00 of the following day to generate a new nighttime period; then, use Monte Carlo sampling to obtain the photovoltaic probability distribution formula expressed by formula (5):

[0024]

[0025] Where: a t and β t are the Beta shape parameters at different times;

[0026] Step 1.3: The probability model of wind power generation adopts a two-parameter Weibull model, and its function expression is:

[0027]

[0028] Where: k is the shape of Weibull distribution; v is the random wind speed; c is the scale parameter of Weibull distribution;

[0029] When the wind speed reaches the cut-in value, the wind turbine starts to work; when the wind speed reaches the rated value, the active output of the generator reaches the maximum value, and then the wind speed continues to increase within a certain range, and its output remains unchanged; when the wind speed reaches the cut-out value, the wind turbine does not work. The relationship between the wind turbine output and wind speed is expressed as:

[0030]

[0031] Where: P wind is the random output of the wind turbine; v is the wind speed; v r is the rated value of wind speed; v ci is the cut-in value of wind speed; v co Wind speed cut-out value; P r is the rated output power of the wind turbine;

[0032] In addition, the reactive power of the wind turbine is expressed by its active power as:

[0033] Q wind =P wind ·tanφ (9)

[0034] Where: φ is the power factor angle;

[0035] Step 1.4: Use normal distribution to describe the fluctuation characteristics of the load. The corresponding expression is as follows:

[0036]

[0037] Where: P L and Q L are respectively the mean values ​​of active power and reactive power of the load; μ PL and μ QL are the standard deviations of active power and reactive power, respectively;

[0038] Step 1.5: Monte Carlo sampling method is used for a random variable X, that is, the output power of the distributed power source. The cumulative distribution function of the output power X is expressed as formula (11):

[0039] Y=F(X) (11)

[0040] Where: X is a random variable;

[0041] The vertical axis represents probability, which is consistent with the numerical range of the random number sequence, both [0,1]. It is known that the random number sequence represents the probability value. Substitute the random number Φ(i) into the inverse function of the probability density function X=F -1 (Y), the distributed power output power sample value X at the i-th sample point is obtained Di , after multiple samplings, a large number of sample values ​​of distributed power output can be obtained.

[0042] In formula (7) of step 1.3, a and b are both constants, calculated by formula (8);

[0043]

[0044] Step 2 is specifically as follows: a large number of wind and solar scenes generated in step 1 are input into the Matpower simulation software for power flow calculation to obtain the voltage data of each node.

[0045] In step 2: the flow equation is calculated as:

[0046]

[0047] Where: W is the system node power injection vector; X is the node voltage vector; Y is the system network parameter; Z is the system branch power flow vector.

[0048] Step 3 is as follows:

[0049] Step 3.1: Obtain the Gaussian mixture model expression GMM (x) is represented as follows:

[0050]

[0051] Where: n is the number of sub-Gaussians; f(x) is the kth sub-Gaussian of the Gaussian mixture probability density function; ω k , μ k and σ k are the weight coefficient, expectation and standard deviation of the kth sub-Gaussian respectively; where ω k It has the following properties:

[0052]

[0053] Step 3.2: Use the EM algorithm to iteratively optimize the model parameters until convergence. The model is trained and used to fit the voltage data and calculate the probability density function for any given voltage value.

[0054] Step 3.2 is as follows:

[0055] (1) Initialization parameters: set the number of iterations of the EM optimization algorithm, the optimization termination condition, and the matrix to be estimated;

[0056] (2) Use the EM algorithm;

[0057] (3) Bayesian selection criterion BIC is used to determine the optimal number of models for the mixed Gaussian distribution. The method of determination is that the number of models corresponding to the minimum BIC value is the optimal number of models; then, voltage samples of different nodes are collected, and the above-mentioned EM algorithm is used to iteratively optimize the model parameters until convergence; the likelihood function is maximized to find the parameter combination that is most likely to generate the observed data; once the model is trained, it is used to calculate the probability density function of any given voltage value to complete the probability distribution estimation.

[0058] The EM algorithm is specifically:

[0059] Step E: Calculate the probability that each data i comes from sub-model k. The expression is as follows:

[0060]

[0061] Step M: Calculate the model parameters of the new iteration. The expression is as follows:

[0062]

[0063] Iterate the E-step and M-step until the optimization termination condition is reached.

[0064] The beneficial effects of the present invention are:

[0065] (1) Compared with the traditional Monte Carlo sampling, the proposed Monte Carlo method for day and night segmentation can directly reduce the demand for computing resources, making the entire probabilistic power flow calculation process more efficient.

[0066] (2) The mixed Gaussian distribution in the method of the present invention can fit voltage probability density functions of any shape. No matter how complex the distribution of voltage data is, the parameters of each Gaussian component can be adjusted to approximate the true distribution, thereby accurately calculating the voltage over-limit probability and voltage comprehensive risk index of each node.

[0067] (3) The method of the present invention explores the voltage over-limit probability and voltage comprehensive risk index of the distribution system under different capacities, proportions and operating conditions of wind power and photovoltaic access. The evaluation results can provide guidance for the operation and planning of distribution networks containing wind power and photovoltaics. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is the Monte Carlo sampling principle diagram in the method of the present invention;

[0069] Figure 2 is a flow chart of voltage over-limit risk assessment in the method of the present invention;

[0070] Figure 3 It is a topological diagram of the IEEE 33 power distribution system including wind and solar in the method of the present invention. DETAILED DESCRIPTION

[0071] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0072] The present invention provides a method for assessing voltage over-limit risk in a wind-solar access power distribution system. Figure 2 As shown, the following steps are included:

[0073] Step 1: First, the irradiance and wind speed in the distributed power probability model are processed by time period, and then Monte Carlo is used to perform day and night segment sampling to generate a large number of wind and solar scenes;

[0074] Step 1 is implemented as follows:

[0075] Step 1.1: In the natural environment, since the intensity of light cannot be controlled by humans, it has great uncertainty, but it is consistent with the Beta distribution within a certain time interval; and the amount of power that a photovoltaic power plant can generate is closely related to the intensity of light in the area. In addition, in the probability model of photovoltaic power station output, the relationship between photovoltaic power station output and light intensity is:

[0076]

[0077] Where: γ max is the maximum value of light intensity; γ is the actual value of light intensity; Γ is the Gamma function; α and β are the shape parameters of Beta distribution.

[0078] In this probability model, the average and standard values ​​of the light intensity required for photovoltaic cells to generate power within a certain period of time are approximately:

[0079]

[0080] Where: μ is the mean; δ 2 is the variance; α and β are the shape parameters of the Beta distribution.

[0081] The output power of the photovoltaic cell is:

[0082]

[0083] Where: P M is the actual output power of the photovoltaic cell; A is the total area of ​​the photovoltaic cell; η is the photoelectric conversion power of the photovoltaic cell; R M is the maximum output power of the photovoltaic cell; γ max is the maximum value of light intensity.

[0084] The probability density of photovoltaic output is obtained from equations (1) to (3):

[0085]

[0086] Step 1.2: As shown in formula (4), the traditional calculation of light intensity is usually based on days / months, and does not take into account the hourly level. At the same time, seasonal factors and weather conditions are prerequisites for determining the photovoltaic light intensity model. This will result in the data for calculating the Beta distribution parameters being inaccurate and unable to simulate the actual local photovoltaic output. Therefore, a day-night segmented Monte Carlo sampling method is proposed. Taking photovoltaic power sources as an example, first, the photovoltaic Beta parameters calculated by the above method at the hourly level are divided into two segments to generate daytime and nighttime periods; the daytime period is 6:00-18:00, and the nighttime period is 0:00-6:00 and 18:00-24:00; then, the nighttime period 18:00-24:00 is recombined with the time period 0:00-6:00 of the next day to generate a new nighttime period; then, Monte Carlo sampling is used in the generated daytime period and nighttime period respectively; therefore, the photovoltaic probability distribution formula is expressed by formula (5):

[0087]

[0088] Where: a t and β t are the Beta shape parameters at different times;

[0089] Step 1.3: Similar to the light intensity, the change of wind speed in the wind farm is also uncertain, and its magnitude is related to the local weather and different time periods of the day. Through the analysis and summary of various data, it can be seen that the average wind speed in general areas obeys the Weibull distribution. In order to make the wind speed data generated by the wind farm probability model consistent with the actual wind speed in most areas of reality, the probability model of wind power generation in this paper adopts a two-parameter Weibull model, and its function expression is:

[0090]

[0091] Where: k is the shape of Weibull distribution; v is the random wind speed; c is the scale parameter of Weibull distribution;

[0092] When the wind speed reaches the cut-in value, the wind turbine starts to work; when the wind speed reaches the rated value, the active output of the generator reaches the maximum value, and then the wind speed continues to increase within a certain range, and its output remains unchanged; when the wind speed reaches the cut-out value, the wind turbine does not work. The relationship between the wind turbine output and wind speed is expressed as:

[0093]

[0094] Where: P wind is the random output of the wind turbine; v is the wind speed; v r is the rated value of wind speed; v ci is the cut-in value of wind speed; v co Wind speed cut-out value; P r is the rated output power of the wind turbine; a and b are both constants, calculated by formula (8);

[0095]

[0096] In addition, the reactive power of the wind turbine is expressed by its active power as:

[0097] Q wind =P wind ·tanφ (9)

[0098] Where: φ is the power factor angle;

[0099] Step 1.4: In addition to the randomness of distributed power sources due to their own characteristics, the load power within a period of time also has volatility. Generally speaking, the fluctuation characteristics of the load are described by normal distribution, and the corresponding expression is as follows:

[0100]

[0101] Where: P L and Q Lare respectively the mean values ​​of active power and reactive power of the load; μ PL and μ QL are the standard deviations of active power and reactive power, respectively.

[0102] Step 1.5: Figure 1 As shown in the figure, the Monte Carlo (MC) sampling method simulates the output distribution of distributed generation and load demand distribution by generating a large number of pseudo-random numbers, and then obtains the risk index reflecting the operating status of the distribution network with the help of probabilistic power flow calculation.

[0103] The pseudo-random number that meets the requirements can be obtained by sampling the established probability model. The process of sampling a random variable X (distributed power output power) using the Monte Carlo sampling method is as follows: Figure 1 As shown, the cumulative distribution function (CDF) of the output power X is expressed as formula (11):

[0104] Y=F(X) (11)

[0105] Where: X is a random variable;

[0106] The vertical axis represents probability, which is consistent with the numerical range of the random number sequence, both [0,1]. It is known that the random number sequence represents the probability value. Substitute the random number Φ(i) into the inverse function of the probability density function X=F -1 (Y), the distributed power output power sample value X at the i-th sample point is obtained Di , after multiple samplings, a large number of sample values ​​of distributed power output can be obtained.

[0107] Step 2: Input a large number of wind and solar scenarios generated in step 1 into the Matpower simulation software for power flow calculation to obtain the voltage data of each node;

[0108] In step 2, power flow calculation is the basis of power system analysis, and its typical calculation power flow equation is:

[0109]

[0110] Where: W is the system node power injection vector; X is the node voltage vector; Y is the system network parameter; Z is the system branch power flow vector.

[0111] Since there are various uncertain factors in the distribution network, such as load fluctuations, fluctuations in renewable energy output, equipment failures, etc., the above situations should be considered when performing power flow calculations. Probabilistic power flow calculations can consider the impact of these factors on the system power flow distribution, thereby more accurately evaluating the stability and reliability of the system.

[0112] Step 3: Use the mixed Gaussian model to fit the voltage data in step 2 to obtain the voltage probability density function. The Gaussian mixture model is equivalent to a tool. As long as there is a large amount of voltage data, the voltage data can be fitted and the voltage probability density function can be calculated.

[0113] Step 3 is implemented as follows:

[0114] In the probability flow, the output node voltage is a random variable that does not completely obey the normal distribution. Its probability distribution can be fitted by the mixed Gaussian model. The mixed Gaussian model is a model that describes the mixed density distribution and is widely used in pattern recognition, cluster analysis and other fields. It is not limited to a specific probability density function form assumption, and any probability density distribution can be approximated by a linear combination of several Gaussian density functions, which can accurately characterize the voltage probability density function.

[0115] Step 3.1: Obtain the Gaussian mixture model expression GMM (x) is represented as follows:

[0116]

[0117] Where: n is the number of sub-Gaussians; f(x) is the kth sub-Gaussian of the Gaussian mixture probability density function; ω k , μ k and σ k are the weight coefficient, expectation and standard deviation of the kth sub-Gaussian respectively; where ω k It has the following properties:

[0118]

[0119] Step 3.2: The more sub-Gaussians the mixed Gaussian model corresponds to, the better the fitting effect, but the amount of calculation will also increase greatly. Therefore, it is necessary to select the appropriate number of sub-Gaussians n according to the probability density distribution of the voltage. Since the Gaussian mixture model has hidden states, this paper uses the maximum likelihood algorithm and the Bayesian selection criterion to determine the parameters of the sub-model of the mixed Gaussian model. The basic idea is to iteratively estimate the model parameters through the re-estimation formula, so that the likelihood function reaches the maximum, which greatly reduces the computational complexity of the maximum likelihood estimation; the EM algorithm (maximum expectation algorithm) is used to iteratively optimize the model parameters until convergence. This process will maximize the likelihood function and find the parameter combination that is most likely to generate the observed data. Once the model is trained, it can be used to fit the voltage data and calculate the probability density function of any given voltage value; the Gaussian mixture model is equivalent to a tool. As long as there is a large amount of voltage data, the voltage data can be fitted and the voltage probability density function can be calculated.

[0120] The specific process is as follows:

[0121] (1) Initialization parameters: set the number of iterations of the EM optimization algorithm, the optimization termination condition, and the matrix to be estimated;

[0122] (2) Using the EM algorithm:

[0123] Step E: Calculate the probability that each data i comes from sub-model k. The expression is as follows:

[0124]

[0125] Step M: Calculate the model parameters of the new iteration. The expression is as follows:

[0126]

[0127] Iterate the E step and the M step until the optimization termination condition is reached;

[0128] (3) Bayesian selection criterion (BIC): Use BIC to determine the optimal number of models for the mixed Gaussian distribution. The method of determination is that the number of models corresponding to the minimum BIC value is the optimal number of models. Then, voltage samples of different nodes are collected, and the maximum expectation algorithm (EM algorithm) mentioned above is used to iteratively optimize the model parameters until convergence. This process maximizes the likelihood function, thereby finding the parameter combination that is most likely to generate the observed data. Once the model is trained, it can be used to calculate the probability density function of any given voltage value.

[0129] (4) Obtain the voltage probability density function and complete the probability distribution estimation.

[0130] Step 4: Based on the voltage probability density function in step 3, further integrate the known voltage probability density function to calculate the voltage over-limit probability and voltage comprehensive evaluation index of the node.

[0131] Book Figure 2 The following is a flowchart for voltage over-limit risk assessment. First, obtain irradiance data, wind speed, load and other data, and perform data preprocessing. Based on the probability characteristics of distributed power sources, a large number of wind and solar scenes are generated by Monte Carlo day and night segment sampling. The generated scenes are further used as input to perform power flow calculations on Matpower to obtain a large amount of voltage data at each node. The calculated voltage is fitted with a probability density using a mixed Gaussian model, and the probability density function expression is calculated. The probability of node voltage over-limit and the comprehensive evaluation index of voltage over-limit risk are calculated through the explicit function of each node.

[0132] like Figure 3As shown in the figure, based on the distribution network system, wind farms and photovoltaics are connected at nodes 18 and 33, with maximum outputs of 800kW and 1200kW respectively. Assuming that the power factor of photovoltaic output is 0.98, the load fluctuation obeys the normal distribution, the standard node load is taken as the mean, and the standard deviation is 10% of the mean. The method proposed in this paper is used to evaluate the voltage over-limit risk of the distribution network system.

[0133] The method proposed in this paper is used to sample and generate scenery scenes, and the mean of a large number of scenes is calculated. The speed of MC simulation with day and night segments and the speed of traditional MC simulation are shown in Table 1. When the number of simulations increases, the MC method with day and night segments has an advantage in calculation speed over the traditional MC sampling method.

[0134] Table 1 Comparison of calculation speed of different sampling methods

[0135]

[0136] Verify the distribution network voltage probability distribution based on the mixed Gaussian model, and compare it with Gaussian fitting and maximum entropy fitting to verify the accuracy and effectiveness of the distribution network voltage probability based on the mixed Gaussian model. Because the mixed Gaussian model fitting can capture more complex structures and patterns of data and has greater flexibility, while the maximum entropy model tends to be a simple model, which may ignore some important information in the data.

[0137] (1) For the example of only photovoltaic, wind power, and wind-solar in the system, the voltage over-limit probability and risk index of different access types are shown in Table 2;

[0138] Table 2 Voltage over-limit probability and risk index for different access types

[0139]

[0140] When the access capacity of distributed power sources is the same, consider three situations: only wind power generation is connected to the system, only photovoltaic power generation is connected, and wind and solar power generation are connected at the same time. With the access of distributed power sources, the voltage fluctuation of the system becomes larger, and the voltage over-limit probability and the comprehensive index of voltage over-limit risk are also increasing. Among them, only photovoltaic access is the most serious, and only wind power access is relatively serious. It can be obviously seen that the mixed access effect of wind and solar power sources is significantly better than that of single distributed power source access.

[0141] This is because of the bi-peak characteristics of the load, when photovoltaic power generation is connected, the photovoltaic output is sufficient during the day. When the photovoltaic output is greater than the load power consumption, the voltage will rise, while the photovoltaic output is almost zero at night. The photovoltaic output is lower than the load power consumption, which will cause the voltage to drop, resulting in the situation that the upper limit is higher during the day and the lower limit is lower at night; when wind power generation is connected, excessive output during the day will cause the voltage to exceed the upper limit, while the load power consumption is less at night, so the voltage will also be too high at night; when wind power generation is connected in coordination with photovoltaic power generation, wind and solar resources output at the same time during the day. At night, due to the characteristics of photovoltaic output, photovoltaic power generation cannot provide support, while wind power generation continues to output to maintain system voltage stability. Therefore, when wind power generation is connected in coordination with photovoltaic power generation, wind and solar resources can be fully utilized to reduce the probability of voltage exceeding the limit.

[0142] (2) When wind and solar capacity are added at the same time, the voltage over-limit probability and risk index of different access capacities are shown in Table 3;

[0143] Table 3 Voltage over-limit probability and risk index for different access capacities

[0144]

[0145] In order to consider the impact of photovoltaic and wind power generation on the system voltage over-limit risk of the distribution network after access, and to explore the impact of different capacity access on the distribution network, the maximum access capacity of the two is changed. With the increase of wind power generation and photovoltaic power generation capacity, the probability trend of distribution network voltage over-limit is generally increasing. During the day, with the increase of wind power generation and photovoltaic power generation capacity, the system over-limit risk increases. At night, with the increase of wind power generation capacity, the over-limit risk first decreases and then increases.

[0146] (3) Changing the ratio of wind and solar power. At different nodes, the voltage over-limit probability and risk index of different ratios of wind and solar power connected at different nodes are shown in Table 4;

[0147] Table 4 Voltage over-limit probability and risk index of different proportions of wind and solar power connected to different nodes

[0148]

[0149] The capacity change of distributed energy access will cause system voltage fluctuations. From the above example, it can be seen that when the capacity of distributed energy is too large, the probability of voltage over-limit is high. In order to make full use of photovoltaic and wind energy, 18 nodes and 33 nodes are selected to access distributed power sources with a total capacity of 1600kW, of which 18 nodes are connected to wind power and 33 nodes are connected to photovoltaic. Changing the ratio of wind power and photovoltaic power, the total capacity of distributed power sources is given, and the proportion of wind and solar power generation is different, which has a great impact on the voltage of the system. The situation during the day and at night is different. During the day, as the wind power generation capacity increases and the photovoltaic power generation capacity decreases, the risk of system over-limit increases. At night, as the wind power generation capacity increases, the risk of voltage over-limit gradually decreases. This is because when the load power consumption is too high at night, due to the characteristics of photovoltaic power generation not working at night, increasing the capacity of wind power generation can effectively increase the voltage at night and reduce the risk of voltage over-limit at night. Combined with the output characteristics of wind power generation, it can be seen that increasing the capacity of wind power generation and reducing the capacity of photovoltaic power generation can increase the system voltage stability to a certain extent when the voltage drops at night. Selecting different ratios of photovoltaic and wind power generation will affect the voltage over-limit risk of the distribution network. Reasonable selection of photovoltaic access points is of great significance to the stable operation of the distribution network. Therefore, the method of the present invention quantitatively analyzes the impact of wind and solar access to the distribution system under various conditions based on the comprehensive risk index of voltage over-limit. The operating risk of the system is related to the form of wind and solar access. The voltage over-limit probability and comprehensive evaluation index are small when wind and solar complementary access. The risk level of the system is different for different wind and solar access ratios. In addition, when the working conditions change, the voltage over-limit probability and the size of the comprehensive evaluation index in the system will also change accordingly. The evaluation results can provide guidance for the operation and planning of electric distribution systems containing new energy, and have certain engineering application value.

[0150] Example 1

[0151] The method for assessing voltage over-limit risk of wind and solar power access distribution system includes the following steps:

[0152] Step 1: First, the irradiance and wind speed in the distributed power probability model are processed by time period, and then Monte Carlo is used to perform day and night segment sampling to generate a large number of wind and solar scenes;

[0153] Step 2: Perform power flow calculation on a large number of wind and solar scenarios generated in step 1 to obtain voltage data of each node;

[0154] Step 3: Use the mixed Gaussian model to fit the voltage data in step 2 to obtain the voltage probability density function;

[0155] Step 4: Based on the voltage probability density function in step 3, further calculate the voltage over-limit probability and voltage comprehensive evaluation index of the node.

[0156] Example 2

[0157] The method for assessing voltage over-limit risk of wind and solar power access distribution system includes the following steps:

[0158] Step 1: First, the irradiance and wind speed in the distributed power probability model are processed by time period, and then Monte Carlo is used to perform day and night segment sampling to generate a large number of wind and solar scenes;

[0159] Step 1 is as follows:

[0160] Step 1.1: In the probability model of PV power station output, the relationship between PV power station output and light intensity is:

[0161]

[0162] Where: γ max is the maximum value of light intensity; γ is the actual value of light intensity; Γ is the Gamma function; α and β are both shape parameters of Beta distribution;

[0163] In this probability model, the average and standard values ​​of the light intensity required for photovoltaic cells to generate power within a certain period of time are approximately:

[0164]

[0165] Where: μ is the mean; δ 2 is the variance; α and β are the shape parameters of Beta distribution;

[0166] The output power of the photovoltaic cell is:

[0167]

[0168] Where: P M is the actual output power of the photovoltaic cell; A is the total area of ​​the photovoltaic cell; η is the photoelectric conversion power of the photovoltaic cell; R M is the maximum output power of the photovoltaic cell; γ max is the maximum value of light intensity;

[0169] The probability density of photovoltaic output is obtained from equations (1) to (3):

[0170]

[0171] Step 1.2: Divide the photovoltaic beta parameters calculated at the hourly level into two segments, generating daytime and nighttime periods; the daytime period is 6:00-18:00, and the nighttime periods are 0:00-6:00 and 18:00-24:00; then, recombine the nighttime period 18:00-24:00 with the time period 0:00-6:00 of the following day to generate a new nighttime period; then, use Monte Carlo sampling to obtain the photovoltaic probability distribution formula expressed by formula (5):

[0172]

[0173] Where: a t and β t are the Beta shape parameters at different times;

[0174] Step 1.3: The probability model of wind power generation adopts a two-parameter Weibull model, and its function expression is:

[0175]

[0176] Where: k is the shape of Weibull distribution; v is the random wind speed; c is the scale parameter of Weibull distribution;

[0177] When the wind speed reaches the cut-in value, the wind turbine starts to work; when the wind speed reaches the rated value, the active output of the generator reaches the maximum value, and then the wind speed continues to increase within a certain range, and its output remains unchanged; when the wind speed reaches the cut-out value, the wind turbine does not work. The relationship between the wind turbine output and wind speed is expressed as:

[0178]

[0179] Where: P wind is the random output of the wind turbine; v is the wind speed; v r is the rated value of wind speed; v ci is the cut-in value of wind speed; v co Wind speed cut-out value; P r is the rated output power of the wind turbine;

[0180] In addition, the reactive power of the wind turbine is expressed by its active power as:

[0181] Q wind =P wind ·tanφ (9)

[0182] Where: φ is the power factor angle;

[0183] Step 1.4: Use normal distribution to describe the fluctuation characteristics of the load. The corresponding expression is as follows:

[0184]

[0185] Where: P L and Q L are respectively the mean values ​​of active power and reactive power of the load; μ PL and μ QL are the standard deviations of active power and reactive power, respectively;

[0186] Step 1.5: Monte Carlo sampling method is used for a random variable X, that is, the output power of the distributed power source. The cumulative distribution function of the output power X is expressed as formula (11):

[0187] Y=F(X) (11)

[0188] Where: X is a random variable;

[0189] The probability and the random number sequence have the same numerical range, both [0,1]. It is known that the random number sequence represents the probability value. Substitute the random number Φ(i) into the inverse function of the probability density function X=F -1 (Y), the distributed power output power sample value X at the i-th sample point is obtained Di , after multiple samplings, a large number of sample values ​​of distributed power output can be obtained.

[0190] Step 2: Perform power flow calculation on a large number of wind and solar scenarios generated in step 1 to obtain voltage data of each node;

[0191] Step 3: Use the mixed Gaussian model to fit the voltage data in step 2 to obtain the voltage probability density function;

[0192] Step 4: Based on the voltage probability density function in step 3, further calculate the voltage over-limit probability and voltage comprehensive evaluation index of the node.

[0193] Example 3

[0194] The method for assessing voltage over-limit risk of wind and solar power access distribution system includes the following steps:

[0195] Step 1: First, the irradiance and wind speed in the distributed power probability model are processed by time period, and then Monte Carlo is used to perform day and night segment sampling to generate a large number of wind and solar scenes;

[0196] Step 1 is as follows:

[0197] Step 1.1: In the probability model of PV power station output, the relationship between PV power station output and light intensity is:

[0198]

[0199] Where: γ max is the maximum value of light intensity; γ is the actual value of light intensity; Γ is the Gamma function; α and β are both shape parameters of Beta distribution;

[0200] In this probability model, the average and standard values ​​of the light intensity required for photovoltaic cells to generate power within a certain period of time are approximately:

[0201]

[0202] Where: μ is the mean; δ 2 is the variance; α and β are the shape parameters of Beta distribution;

[0203] The output power of the photovoltaic cell is:

[0204]

[0205] Where: P M is the actual output power of the photovoltaic cell; A is the total area of ​​the photovoltaic cell; η is the photoelectric conversion power of the photovoltaic cell; R M is the maximum output power of the photovoltaic cell; γ max is the maximum value of light intensity;

[0206] The probability density of photovoltaic output is obtained from equations (1) to (3):

[0207]

[0208] Step 1.2: Divide the photovoltaic beta parameters calculated at the hourly level into two segments, generating daytime and nighttime periods; the daytime period is 6:00-18:00, and the nighttime periods are 0:00-6:00 and 18:00-24:00; then, recombine the nighttime period 18:00-24:00 with the time period 0:00-6:00 of the following day to generate a new nighttime period; then, use Monte Carlo sampling to obtain the photovoltaic probability distribution formula expressed by formula (5):

[0209]

[0210] Where: a t and β t are the Beta shape parameters at different times;

[0211] Step 1.3: The probability model of wind power generation adopts a two-parameter Weibull model, and its function expression is:

[0212]

[0213] Where: k is the shape of Weibull distribution; v is the random wind speed; c is the scale parameter of Weibull distribution;

[0214] When the wind speed reaches the cut-in value, the wind turbine starts to work; when the wind speed reaches the rated value, the active output of the generator reaches the maximum value, and then the wind speed continues to increase within a certain range, and its output remains unchanged; when the wind speed reaches the cut-out value, the wind turbine does not work. The relationship between the wind turbine output and wind speed is expressed as:

[0215]

[0216] Where: P wind is the random output of the wind turbine; v is the wind speed; v r is the rated value of wind speed; v ci is the cut-in value of wind speed; v co Wind speed cut-out value; P r is the rated output power of the wind turbine;

[0217] In addition, the reactive power of the wind turbine is expressed by its active power as:

[0218] Q wind =P wind ·tanφ (9)

[0219] Where: φ is the power factor angle;

[0220] In formula (7) of step 1.3, a and b are both constants, calculated by formula (8);

[0221]

[0222] Step 1.4: Use normal distribution to describe the fluctuation characteristics of the load. The corresponding expression is as follows:

[0223]

[0224] Where: P L and Q L are respectively the mean values ​​of active power and reactive power of the load; μ PL and μ QL are the standard deviations of active power and reactive power, respectively;

[0225] Step 1.5: Monte Carlo sampling method is used for a random variable X, that is, the output power of the distributed power source. The cumulative distribution function of the output power X is expressed as formula (11):

[0226] Y=F(X) (11)

[0227] Where: X is a random variable;

[0228] The probability and the random number sequence have the same numerical range, both [0,1]. It is known that the random number sequence represents the probability value. Substitute the random number Φ(i) into the inverse function of the probability density function X=F -1 (Y), the distributed power output power sample value X at the i-th sample point is obtained Di , after multiple samplings, a large number of sample values ​​of distributed power output can be obtained.

[0229] Step 2: Perform power flow calculation on a large number of wind and solar scenarios generated in step 1 to obtain voltage data of each node;

[0230] Step 3: Use the mixed Gaussian model to fit the voltage data in step 2 to obtain the voltage probability density function;

[0231] Step 4: Based on the voltage probability density function in step 3, further calculate the voltage over-limit probability and voltage comprehensive evaluation index of the node.

[0232] Example 4

[0233] The method for assessing voltage over-limit risk of wind and solar power access distribution system includes the following steps:

[0234] Step 1: First, the irradiance and wind speed in the distributed power probability model are processed by time period, and then Monte Carlo is used to perform day and night segment sampling to generate a large number of wind and solar scenes;

[0235] Step 1 is as follows:

[0236] Step 1.1: In the probability model of PV power station output, the relationship between PV power station output and light intensity is:

[0237]

[0238] Where: γ max is the maximum value of light intensity; γ is the actual value of light intensity; Γ is the Gamma function; α and β are both shape parameters of Beta distribution;

[0239] In this probability model, the average and standard values ​​of the light intensity required for photovoltaic cells to generate power within a certain period of time are approximately:

[0240]

[0241] Where: μ is the mean; δ 2 is the variance; α and β are the shape parameters of Beta distribution;

[0242] The output power of the photovoltaic cell is:

[0243]

[0244] Where: P M is the actual output power of the photovoltaic cell; A is the total area of ​​the photovoltaic cell; η is the photoelectric conversion power of the photovoltaic cell; R M is the maximum output power of the photovoltaic cell; γ max is the maximum value of light intensity;

[0245] The probability density of photovoltaic output is obtained from equations (1) to (3):

[0246]

[0247] Step 1.2: Divide the photovoltaic beta parameters calculated at the hourly level into two segments, generating daytime and nighttime periods; the daytime period is 6:00-18:00, and the nighttime periods are 0:00-6:00 and 18:00-24:00; then, recombine the nighttime period 18:00-24:00 with the time period 0:00-6:00 of the following day to generate a new nighttime period; then, use Monte Carlo sampling to obtain the photovoltaic probability distribution formula expressed by formula (5):

[0248]

[0249] Where: a t and β t are the Beta shape parameters at different times;

[0250] Step 1.3: The probability model of wind power generation adopts a two-parameter Weibull model, and its function expression is:

[0251]

[0252] Where: k is the shape of Weibull distribution; v is the random wind speed; c is the scale parameter of Weibull distribution;

[0253] When the wind speed reaches the cut-in value, the wind turbine starts to work; when the wind speed reaches the rated value, the active output of the generator reaches the maximum value, and then the wind speed continues to increase within a certain range, and its output remains unchanged; when the wind speed reaches the cut-out value, the wind turbine does not work. The relationship between the wind turbine output and wind speed is expressed as:

[0254]

[0255] Where: P wind is the random output of the wind turbine; v is the wind speed; v r is the rated value of wind speed; v ci is the cut-in value of wind speed; v co Wind speed cut-out value; Pr is the rated output power of the wind turbine;

[0256] In addition, the reactive power of the wind turbine is expressed by its active power as:

[0257] Q wind =P wind ·tanφ (9)

[0258] Where: φ is the power factor angle;

[0259] In formula (7) of step 1.3, a and b are both constants, calculated by formula (8);

[0260]

[0261] Step 1.4: Use normal distribution to describe the fluctuation characteristics of the load. The corresponding expression is as follows:

[0262]

[0263] Where: P L and Q L are respectively the mean values ​​of active power and reactive power of the load; μ PL and μ QL are the standard deviations of active power and reactive power, respectively;

[0264] Step 1.5: Monte Carlo sampling method is used for a random variable X, that is, the output power of the distributed power source. The cumulative distribution function of the output power X is expressed as formula (11):

[0265] Y=F(X) (11)

[0266] Where: X is a random variable;

[0267] The probability and the random number sequence have the same numerical range, both [0,1]. It is known that the random number sequence represents the probability value. Substitute the random number Φ(i) into the inverse function of the probability density function X=F -1 (Y), the distributed power output power sample value X at the i-th sample point is obtained Di , after multiple samplings, a large number of sample values ​​of distributed power output can be obtained.

[0268] Step 2: Perform power flow calculation on a large number of wind and solar scenarios generated in step 1 to obtain voltage data of each node;

[0269] Step 2 is specifically as follows: a large number of wind and solar scenes generated in step 1 are input into the Matpower simulation software for power flow calculation to obtain the voltage data of each node.

[0270] Step 3: Use the mixed Gaussian model to fit the voltage data in step 2 to obtain the voltage probability density function;

[0271] Step 4: Based on the voltage probability density function in step 3, further calculate the voltage over-limit probability and voltage comprehensive evaluation index of the node.

Claims

1. A method for assessing voltage over-limit risk in wind and solar power access distribution systems, characterized in that: The following steps are involved: Step 1: First, the irradiance and wind speed in the distributed power probability model are processed by time period, and then Monte Carlo is used to perform day and night segment sampling to generate a large number of wind and solar scenes; Step 2: Perform power flow calculation on a large number of wind and solar scenarios generated in step 1 to obtain voltage data of each node; Step 3: Use the mixed Gaussian model to fit the voltage data in step 2 to obtain the voltage probability density function; Step 4: Based on the voltage probability density function in step 3, further calculate the voltage over-limit probability and voltage comprehensive evaluation index of the node.

2. The method for assessing voltage over-limit risk of wind-solar access power distribution system according to claim 1 is characterized in that: Step 1 is as follows: Step 1.1: In the probability model of PV power station output, the relationship between PV power station output and light intensity is: Where: γ max is the maximum value of light intensity; γ is the actual value of light intensity; Γ is the Gamma function; α and β are both shape parameters of Beta distribution; In this probability model, the average and standard values ​​of the light intensity required for photovoltaic cells to generate power within a certain period of time are approximately: Where: μ is the mean; δ 2 is the variance; α and β are the shape parameters of Beta distribution; The output power of the photovoltaic cell is: Where: P M is the actual output power of the photovoltaic cell; A is the total area of ​​the photovoltaic cell; η is the photoelectric conversion power of the photovoltaic cell; R M is the maximum output power of the photovoltaic cell; γ max is the maximum value of light intensity; The probability density of photovoltaic output is obtained from equations (1) to (3): Step 1.2: Divide the photovoltaic beta parameters calculated at the hourly level into two segments, generating daytime and nighttime periods; the daytime period is 6:00-18:00, and the nighttime periods are 0:00-6:00 and 18:00-24:00; then, recombine the nighttime period 18:00-24:00 with the time period 0:00-6:00 of the following day to generate a new nighttime period; then, use Monte Carlo sampling to obtain the photovoltaic probability distribution formula expressed by formula (5): Where: a t and β t are the Beta shape parameters at different times; Step 1.3: The probability model of wind power generation adopts a two-parameter Weibull model, and its function expression is: Where: k is the shape of Weibull distribution; v is the random wind speed; c is the scale parameter of Weibull distribution; When the wind speed reaches the cut-in value, the wind turbine starts to work; when the wind speed reaches the rated value, the active output of the generator reaches the maximum value, and then the wind speed continues to increase within a certain range, and its output remains unchanged; when the wind speed reaches the cut-out value, the wind turbine does not work. The relationship between the wind turbine output and wind speed is expressed as: Where: P wind is the random output of the wind turbine; v is the wind speed; v r is the rated value of wind speed; v ci is the cut-in value of wind speed; v co Wind speed cut-out value; P r is the rated output power of the wind turbine; In addition, the reactive power of the wind turbine is expressed by its active power as: Q wind =P wind ·tanφ (9) Where: φ is the power factor angle; Step 1.4: Use normal distribution to describe the fluctuation characteristics of the load. The corresponding expression is as follows: Where: P L and Q L are respectively the mean values ​​of active power and reactive power of the load; μ PL and μ QL are the standard deviations of active power and reactive power, respectively; Step 1.5: Monte Carlo sampling method is used for a random variable X, that is, the output power of the distributed power source. The cumulative distribution function of the output power X is expressed as formula (11): Y=F(X) (11) Where: X is a random variable; The probability and the numerical range of the random number sequence are consistent, both are [0,1]. It is known that the random number sequence represents the probability value. Substitute the random number Φ(i) into the inverse function of the probability density function X=F -1 (Y), the distributed power output power sample value X at the i-th sample point is obtained Di , after multiple samplings, a large number of sample values ​​of distributed power output can be obtained.

3. The method for assessing voltage over-limit risk of wind-solar access power distribution system according to claim 2 is characterized in that: In formula (7) of step 1.3, a and b are both constants, calculated by formula (8); 4. The method for assessing voltage over-limit risk of wind-solar access power distribution system according to claim 2 is characterized in that: Step 2 is specifically as follows: a large number of wind and solar scenes generated in step 1 are input into the Matpower simulation software for power flow calculation to obtain the voltage data of each node.

5. The method for assessing voltage over-limit risk of wind-solar access power distribution system according to claim 4 is characterized in that: In step 2: the flow equation is calculated as: Where: W is the system node power injection vector; X is the node voltage vector; Y is the system network parameter; Z is the system branch power flow vector.

6. The method for assessing voltage over-limit risk of wind-solar access power distribution system according to claim 1 is characterized in that: Step 3 is as follows: Step 3.1: Obtain the Gaussian mixture model expression GMM (x) is represented as follows: Where: n is the number of sub-Gaussians; f(x) is the kth sub-Gaussian of the Gaussian mixture probability density function; ω k , μ k and σ k are the weight coefficient, expectation and standard deviation of the kth sub-Gaussian respectively; where ω k It has the following properties: Step 3.2: Use the EM algorithm to iteratively optimize the model parameters until convergence. The model is trained and used to fit the voltage data and calculate the probability density function for any given voltage value.

7. The method for assessing voltage over-limit risk of wind-solar access power distribution system according to claim 6 is characterized in that: Step 3.2 is as follows: (1) Initialization parameters: set the number of iterations of the EM optimization algorithm, the optimization termination condition, and the matrix to be estimated; (2) Use the EM algorithm; (3) Bayesian selection criterion BIC is used to determine the optimal number of models for the mixed Gaussian distribution. The method of determination is that the number of models corresponding to the minimum BIC value is the optimal number of models; then, voltage samples of different nodes are collected, and the above-mentioned EM algorithm is used to iteratively optimize the model parameters until convergence; the likelihood function is maximized to find the parameter combination that is most likely to generate the observed data; once the model is trained, it is used to calculate the probability density function of any given voltage value to complete the probability distribution estimation.

8. The method for assessing voltage over-limit risk of wind-solar access power distribution system according to claim 7 is characterized in that: The EM algorithm is specifically: Step E: Calculate the probability that each data i comes from sub-model k. The expression is as follows: Step M: Calculate the model parameters of the new iteration. The expression is as follows: Iterate the E-step and M-step until the optimization termination condition is reached.

Images