Power transmission line dynamic capacity increasing method based on Bayesian optimization random forest online training

Abstract

The invention discloses a power transmission line dynamic capacity increasing method based on Bayesian optimization random forest online training, and relates to the field of power transmission line detection. The invention aims to solve the problems that a power transmission line heat balance equation model in the existing power transmission line dynamic capacity increasing technology is fixed and lacks pertinence, and the accuracy of power transmission line dynamic capacity increasing is influenced. Weather forecast data and a conductor current value are substituted into a random forest model to calculate conductor temperature until the temperature is higher than 70 DEG C, an I-T curve is constructed, if not, the conductor current value is updated, if the updated value is smaller than the maximum value, the conductor temperature is calculated through the random forest model, and if not, the conductor temperature is calculated through a power transmission line heat balance model. Otherwise, updating the current value of the conductor, and calculating the temperature of the conductor by using a heat balance model of the power transmission line; and calculating and selecting the minimum value of the current-carrying capacity limit based on the I-T curve to calculate a capacity prediction result, and realizing dynamic capacity increase of the power transmission line by using the capacity prediction result.

Description

Technical Field

[0001] The present invention belongs to the field of transmission line detection. Background Art

[0002] With the continuous increase of power consumption load and new energy penetration rate, the problem of insufficient transmission line capacity has become increasingly serious. The transmission line capacity is related to meteorological conditions. In engineering, extremely severe meteorological conditions are often assumed to calculate the transmission line capacity, resulting in waste of transmission line capacity. In order to fully exploit the transmission potential of transmission lines, existing transmission line dynamic capacity increase technologies often calculate the transmission line capacity based on meteorological information and the heat balance equation. However, the heat balance equation model of the transmission line is fixed and lacks pertinence, making it difficult to accurately reflect the relationship among meteorology, conductor current, and conductor temperature under actual operating conditions, which affects the accuracy of transmission line dynamic capacity increase. Summary of the Invention

[0003] The present invention aims to solve the problem that existing transmission line dynamic capacity increase technologies often calculate the transmission line capacity based on meteorological information and the heat balance equation, but the heat balance equation model of the transmission line is fixed and lacks pertinence, making it difficult to accurately reflect the relationship among meteorology, conductor current, and conductor temperature under actual operating conditions, which affects the accuracy of transmission line dynamic capacity increase. Now, a transmission line dynamic capacity increase method based on online training of Bayesian optimization random forest is provided.

[0004] The transmission line dynamic capacity increase method based on online training of Bayesian optimization random forest includes:

[0005] Substitute meteorological forecast data and conductor current values into the random forest model to calculate the conductor temperature. The initial value of the conductor current value is the minimum conductor current I used when training the random forest model tmin , and the meteorological forecast data includes: wind speed, wind direction, solar irradiance, and ambient temperature;

[0006] If the conductor temperature calculated by the random forest model is higher than 70°C, then construct an I-T curve using the conductor current value and the corresponding conductor temperature. Otherwise, update the conductor current value with the first step length. The first step length is (I tmax - I tmin ) / 10, where I tmax is the maximum conductor current used when training the random forest model. The input of the random forest model is meteorological forecast data and conductor current, and the output is conductor temperature;

[0007] If the conductor current value updated with the first step length is less than the maximum conductor current, then recalculate the conductor temperature using the random forest model based on the conductor current value updated with the first step length. Otherwise, calculate the conductor temperature according to the transmission line heat balance model;

[0008] If the conductor temperature calculated by the transmission line thermal balance model is higher than 70 °C, an I-T curve is constructed using the conductor current value and the corresponding conductor temperature; otherwise, the conductor current value is updated with the second step size, and the conductor temperature is recalculated using the transmission line thermal balance model based on the conductor current value updated with the second step size. The second step size is 50 A.

[0009] Based on the I-T curve, the current-carrying capacity limit corresponding to each conductor current value is calculated, the minimum value of the current-carrying capacity limit is selected to calculate the capacity prediction result, and the dynamic capacity increase of the transmission line is realized using this capacity prediction result.

[0010] Further, calculating the conductor temperature according to the transmission line thermal balance model includes:

[0011] T = g(v, θ, I s , T A , I) + T k - g(v, θ, I s , T A , I tmax ),

[0012] where T is the conductor temperature, g() is the transmission line thermal balance model, v is the wind speed, θ is the wind direction, I s is the solar irradiance, T A is the ambient temperature, I is the conductor current, and T k is the conductor temperature calculated by the random forest model when the conductor current takes I tmax .

[0013] Further, the expression of the transmission line thermal balance model g() is:

[0014] a 4 T 4 + a 3 T 3 + a 2 T 2 + a 1 T + a 0 = 0,

[0015] a 4 = πeSD,

[0016] a 3 = πeSD*(4*273),

[0017] a 2 = πeSD*(6*273 2 ),

[0018] a 1 = πeSD*(4*273 3 ) + 9.92(VD) 0.485-I 2 Rα,

[0019] a 0 = πeSD[273 4 -(T A + 273) 4 -I 2 R(1 - 20α)-a s I s D - 9.92T A (VD) 0.485 ,

[0020] where e is the heat dissipation coefficient, S is the Stefan-Boltzmann constant, D is the outer diameter of the wire, R is the AC resistance of the wire per unit length, α is the temperature coefficient of resistance, and a s is the heat absorption coefficient,

[0021] The expression for the intermediate variable V is:

[0022]

[0023] where, is the power factor, is the minimum value of the angle formed by the wind direction and the lines on both sides of the temperature measurement point.

[0024] Furthermore, constructing the I-T curve using the wire current value and the corresponding wire temperature includes:

[0025] Plotting the wire current and the wire temperature as a curve with the current on the abscissa and the temperature on the ordinate, and then translating the curve upward by ΔT α units to obtain the I-T curve;

[0026] The ΔT α is the lower 99% quantile of the wire temperature error.

[0027] Furthermore, the lower 99% quantile of the above-mentioned wire temperature error is obtained by the following formula:

[0028] P(T E < ΔT α ) = 0.99,

[0029] where T E is the wire temperature error in the test set, that is, the true value of the wire temperature minus the predicted value of the random forest model. P(T E < ΔT α ) represents the probability of T E < ΔT α .

[0030] Furthermore, calculating the current-carrying capacity limit corresponding to each wire current value based on the I-T curve includes:

[0031] Calculate the current-carrying capacity limit according to the following formula:

[0032]

[0033] Wherein, I DTR is the current-carrying capacity limit, I i and I i-1 are the wire current values at the i-th and (i - 1)-th iterations respectively, T i and T i-1 are the wire temperatures at the i-th and (i - 1)-th iterations respectively.

[0034] Furthermore, the above-mentioned selection of the minimum value of the current-carrying capacity limit to calculate the capacity prediction result includes:

[0035] Calculate the capacity prediction result according to the following formula:

[0036]

[0037] Wherein, P max is the capacity prediction result, U is the rated voltage of the transmission line, I DTR_min is the minimum value of the current-carrying capacity limit, is the power factor, is the minimum value of the angle formed by the wind direction and the lines on both sides of the temperature measurement point.

[0038] Furthermore, the above-mentioned construction method of the random forest model is as follows:

[0039] Read historical weather forecast data, the wire current of the transmission line, and the wire temperature of the transmission line, filter abnormal data and supplement missing data to obtain a data set;

[0040] Group the data in the data set according to the season, meteorological lead time, and the number of temperature measurement poles;

[0041] Based on the data set, use the Bayesian optimization algorithm to find the optimal hyperparameters, and construct a random forest model based on the optimal hyperparameters.

[0042] Furthermore, the above-mentioned filtering of abnormal data includes:

[0043] Judge whether each element data in the weather forecast data exceeds its value range. If it exceeds, it is regarded as abnormal and deleted;

[0044] The value ranges of the above-mentioned each element data are as follows:

[0045] The value range of the wind speed v is [0, v 5max , where v 5max is the maximum wind speed of the temperature measurement poles in the past 5 years;

[0046] The value range of the wind direction θ is [0, π). The wind direction is 0° with respect to the true north, and the positive direction is clockwise. When converted to radians, if θ ≥ π, then θ minus π;

[0047] Solar irradiance I s has a value range of [0, I s5max , where I s5max is the maximum solar irradiance of the temperature measurement tower in the past 5 years;

[0048] Ambient temperature T A has a value range of [T A5min , T A5max , where T A5min and T A5max are respectively the minimum and maximum ambient temperatures of the temperature measurement tower in the past 5 years.

[0049] Furthermore, the above meteorological forecast data are the hourly forecast values for the next 6 hours of the current hour, and the conductor current and conductor temperature of the transmission line are the measured average values of the current hour.

[0050] The transmission line dynamic capacity increase method based on Bayesian optimization random forest online training according to the present invention constructs a Bayesian optimization random forest model based on the historical operation data of the transmission line, and combines this model with the transmission line heat balance equation to obtain a model for transmission line dynamic capacity increase that can more truly reflect the on-site operation data, improving the accuracy of the transmission line dynamic capacity increase result. Brief Description of the Drawings

[0051] Figure 1 is a flowchart of the transmission line dynamic capacity increase method based on Bayesian optimization random forest online training. Detailed Embodiments

[0052] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention. It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0053] Refer to Figure 1 to specifically describe this embodiment. The transmission line dynamic capacity increase method based on Bayesian optimization random forest online training described in this embodiment includes:

[0054] I. Data Processing

[0055] 1. Read the meteorological forecast data, the conductor current of the transmission line, and the conductor temperature of the transmission line in real time. The meteorological forecast data are the hourly forecast values for the next 6 hours, and the meteorological forecast data include: wind speed v, wind direction θ, solar irradiance I s and ambient temperature T A . The conductor current and conductor temperature of the transmission line are the measured average values for the current hour.

[0056] 2. Determine whether any element in the meteorological forecast data of each temperature-measuring tower exceeds the allowable value range. If so, the meteorological data is regarded as abnormal. The specific judgment criteria are as follows:

[0057] The value range of wind speed v is [0, v 5max , where v 5max is the maximum wind speed of this temperature-measuring tower in the past 5 years;

[0058] The value range of wind direction θ is [0, π), with the due north as 0°, and the clockwise direction as the positive direction, converted to radians. When θ ≥ π, θ minus π;

[0059] The solar irradiance I s has a value range of [0, I s5max , where I s5max is the maximum solar irradiance of this temperature-measuring tower in the past 5 years;

[0060] The ambient temperature T A has a value range of [T A5min , T A5max , where T A5min and T A5max are the minimum and maximum ambient temperatures of this temperature-measuring tower in the past 5 years, respectively.

[0061] Missing data means that if any item of data is missing, the entire data entry is considered missing.

[0062] Filter out abnormal data and supplement missing data.

[0063] 3. Group all the processed data according to the season, the meteorological lead prediction time (abbreviated as the lead time), and the number of temperature-measuring towers. Suppose there are 4 temperature-measuring towers in total. Since the meteorological forecast data is for the next 6 hours, there are 24 data entries in total; since there are four seasons, there are 4×6×4 = 96 groups in total. Determine whether the number of data groups in the dataset is greater than 5000. If it is greater than 5000, delete the oldest group of data.

[0064] II. Model construction

[0065] The dataset is randomly divided into a training set, a validation set, and a test set in a ratio of 3:1:1. The Bayesian optimization algorithm is used to find the optimal hyperparameters. Among them, the parameters of Bayesian optimization are set as follows: the search range of the minimum number of samples in leaf nodes is [1, 10], the search range of the number of decision trees is [1, 100], the upper limit of the number of iterations is 50, and parallel computing is prohibited. Taking the root mean square error of the validation set as the objective function, the Bayesian optimization process is executed, and the optimal hyperparameters are recorded.

[0066] A random forest model is constructed based on the optimal hyperparameters, where the meteorological forecast data and the wire current are input variables, and the wire temperature is the output variable.

[0067] Based on the constructed random forest model, the lower 99% quantile ΔT of the wire temperature error in the test set is calculated by the following formula α :

[0068] P(T E <ΔT α ) = 0.99,

[0069] where T E is the wire temperature error in the test set, that is, the true value of the wire temperature minus the predicted value of the random forest model. P(T E <ΔT α ) represents the probability that T E <ΔT α .

[0070] III. Capacity Prediction

[0071] Initialization: Read the minimum value I tmin and the maximum value I tmax of the wire current in the training set of each group of datasets, and take the minimum value I tmin of the wire current as the initial value of the wire current;

[0072] S1: Substitute the meteorological forecast data and the wire current value I i at the i-th iteration into the random forest model, and calculate the wire temperature T i at the i-th iteration;

[0073] S2: Determine whether the wire temperature T i obtained in S1 is higher than 70 °C. If so, execute S8; otherwise, execute S3:

[0074] S3: Update I tmax with a step size of (I tmin -I i+1 =I i +[(I tmax -I tmin ) / 10], and then execute S4;

[0075] S4: Determine the wire current value I i+1 Whether it is less than the maximum wire current I tmax , if yes, set i = i + 1, then return to S1, otherwise set i = i + 1, then execute S5;

[0076] S5: Calculate the wire temperature T according to the following formula i , then execute S6:

[0077] T i = g(v, θ, I s , T A , I i ) + T k - g(v, θ, I s , T A , I tmax ),

[0078] where, T k is the wire temperature calculated by the random forest model when I i = I tmax ;

[0079] g() is the heat balance model of the transmission line, which is a quartic equation of one variable about the wire temperature T i , and the expression is:

[0080] a 4 T i 4 + a 3 T i 3 + a 2 T i 2 + a 1 T i + a 0 = 0,

[0081] where, a 4 = πeSD,

[0082] a 3 = πeSD*(4*273),

[0083] a 2 = πeSD*(6*273 2 ),

[0084] a 1 = πeSD*(4*273 3 ) + 9.92(VD) 0.485 - I 2 Rα,

[0085] a 0 = πeSD[273 4-(T A +273) 4 -I 2 R(1 - 20α)-a s I s D - 9.92T A (VD) 0.485 ,

[0086] where e is the heat dissipation coefficient, S is the Stefan - Boltzmann constant, D is the outer diameter of the wire, I is the wire current, R is the AC resistance of the wire per unit length, α is the temperature coefficient of resistance, a s is the heat absorption coefficient,

[0087] The intermediate variable V is:

[0088] is the power factor, is the minimum value of the angle formed by the wind direction and the lines on both sides of the temperature measurement point.

[0089] S6: Determine whether the wire temperature T i obtained in S5 is higher than 70 °C. If yes, execute S8; otherwise, execute S7;

[0090] S7: Update I in steps of 50A i+1 = I i + 50, make i = i + 1 and return to S5;

[0091] S8: Plot the wire current and wire temperature at each iteration as a curve with the current as the abscissa and the temperature as the ordinate, and then translate this curve upward by ΔT α units to obtain the I - T curve, and then execute S9;

[0092] S9: Based on the I - T curve obtained in S8, calculate the current - carrying capacity limit according to the following formula, and then execute S10,

[0093]

[0094] where I i-1 is the wire current value at the (i - 1) - th iteration, T i-1 is the wire temperature at the (i - 1) - th iteration;

[0095] S10: Calculate the capacity prediction result P max :

[0096]

[0097] where U is the rated voltage of the transmission line, I DTR_min is the minimum value of the current - carrying capacity limit.

[0098] IV. Error Evaluation

[0099] Read the real-time conductor current and the conductor temperatures of each temperature-measuring tower, substitute the real-time conductor current into each group of I-T curves respectively, calculate the conductor temperature based on the linear interpolation method, and calculate the conductor temperature error by taking the difference between the actual conductor temperature and the calculated value.

[0100] Although the present invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the present invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed, as long as they do not depart from the spirit and scope of the present invention as defined by the appended claims. It should be understood that the different dependent claims and the features described herein can be combined in a manner different from that described in the original claims. It should also be understood that the features described in connection with a single embodiment can be used in other described embodiments.

Claims

1. A method for dynamic capacity expansion of power transmission lines based on online training of Bayesian optimization random forest, characterized in that: include: Substitute the weather forecast data and the conductor current value into the random forest model to calculate the conductor temperature. The initial value of the conductor current value is the minimum conductor current value I used when training the random forest model. tmin , the weather forecast data includes: wind speed, wind direction, solar irradiance and ambient temperature; If the wire temperature calculated by the random forest model is higher than 70°C, the IT curve is constructed using the wire current value and the corresponding wire temperature. Otherwise, the wire current value is updated with the first step length, and the first step length is (I tmax -I tmin ) / 10,I tmax The maximum value of the conductor current used when training the random forest model, the input of the random forest model is the weather forecast data and the conductor current, and the output is the conductor temperature; If the conductor current value updated by the first step is less than the maximum conductor current, the conductor temperature is calculated again using the random forest model based on the conductor current value updated by the first step, otherwise the conductor temperature is calculated according to the transmission line thermal balance model; If the conductor temperature calculated by the transmission line thermal balance model is higher than 70°C, the IT curve is constructed using the conductor current value and the corresponding conductor temperature; otherwise, the conductor current value is updated with a second step length, and the conductor temperature is recalculated using the transmission line thermal balance model based on the conductor current value updated with the second step length, and the second step length is 50A; The current carrying capacity limit corresponding to each conductor current value is calculated based on the IT curve, the minimum value of the current carrying capacity limit is selected to calculate the capacity prediction result, and the capacity prediction result is used to realize dynamic capacity increase of the transmission line.

2. The method for dynamic capacity increase of power transmission lines based on online training of Bayesian optimization random forest according to claim 1 is characterized in that: The calculating the conductor temperature according to the transmission line thermal balance model comprises: T=g(v,θ,I s ,T A ,I)+T k -g(v,θ,I s ,T A ,I tmax ), Where T is the conductor temperature, g() is the thermal balance model of the transmission line, v is the wind speed, θ is the wind direction, I s is the solar irradiance, T A is the ambient temperature, I is the conductor current, T k When the conductor current is I tmax Wire temperature calculated by the random forest model at .

3. The method for dynamic capacity increase of power transmission lines based on online training of Bayesian optimization random forest according to claim 2 is characterized in that: The expression of the transmission line heat balance model g() is: a4T 4 +a3T 3 +a2T 2 +a1T+a0=0, a4=πeSD, a3=πeSD*(4*273), <h2 style=";text-align:left;direction:ltr">a2 = πeSD*(6*273<h2 style=";text-align:left;direction:ltr"> 2 <h2 style=";text-align:left;direction:ltr"> ), a1=πeSD*(4*273 3 )+9.92(VD) 0.485 -I 2 Rα, a0=πeSD[273 4 -(T A +273) 4 ]-I 2 R(1-20α)-a s I s D-9.92T A (VD) 0.485 , Where, e is the heat dissipation coefficient, S is the Stefan Boltzmann constant, D is the outer diameter of the wire, R is the AC resistance of the wire per unit length, α is the resistance temperature coefficient, and a s Heat absorption coefficient, The expression of the intermediate variable V is: in, is the power factor, It is the minimum angle between the wind direction and the lines on both sides of the temperature measurement point.

4. The method for dynamic capacity increase of power transmission lines based on online training of Bayesian optimization random forest according to claim 1, characterized in that: The IT curve is constructed by using the conductor current value and the corresponding conductor temperature, including: Plot the conductor current and conductor temperature into a curve with current as the horizontal axis and temperature as the vertical axis, and then shift the curve upward by ΔT α units, and obtain the IT curve; The ΔT α is the lower 99% quantile of the wire temperature error.

5. The method for dynamic capacity increase of power transmission lines based on online training of Bayesian optimization random forest according to claim 4 is characterized in that: The lower 99% quantile of the wire temperature error is obtained by: P(T E <ΔT α )=0.99, Among them, T E is the conductor temperature error in the test set, that is, the actual conductor temperature minus the predicted value of the random forest model. E <ΔT α ) indicates T E <ΔT α probability.

6. The method for dynamic capacity increase of power transmission lines based on online training of Bayesian optimization random forest according to claim 1, characterized in that: The calculating the current carrying capacity limit corresponding to each conductor current value based on the IT curve includes: Calculate the current carrying capacity limit according to the following formula: Among them, I DTR is the current carrying capacity limit, I i and I i-1 are the conductor current values ​​at the i-th and i-1-th iterations, respectively, T i and T i-1 are the wire temperatures at the i-th and i-1-th iterations respectively.

7. The method for dynamic capacity increase of power transmission lines based on online training of Bayesian optimization random forest according to claim 1, characterized in that: The method of selecting the minimum value of the current carrying capacity limit to calculate the capacity prediction result includes: The capacity forecast is calculated using the following formula: Among them, P max is the capacity prediction result, U is the rated voltage of the transmission line, I DTR_min is the minimum value of the current carrying capacity limit, is the power factor, It is the minimum angle between the wind direction and the lines on both sides of the temperature measurement point.

8. The method for dynamic capacity increase of power transmission lines based on online training of Bayesian optimization random forest according to claim 1, characterized in that: The random forest model is constructed as follows: Read historical weather forecast data, conductor current of transmission lines, and conductor temperature of transmission lines, filter abnormal data, and supplement missing data to obtain a data set; The data in the dataset are grouped according to season, weather advance prediction time and number of temperature measuring towers; Based on the data set, the Bayesian optimization algorithm is used to find the optimal hyperparameters, and a random forest model is built based on the optimal hyperparameters.

9. The method for dynamic capacity increase of power transmission lines based on online training of Bayesian optimization random forest according to claim 8, characterized in that: The filtering of abnormal data includes: Determine whether the data of each element in the weather forecast data exceeds its value range. If it exceeds, it will be considered abnormal and deleted; The value ranges of the data elements are as follows: The value range of wind speed v is [0,v 5max ],v 5max The maximum wind speed of the tower measured in the past five years; The value range of wind direction θ is [0,π), with due north as 0° and clockwise as the positive direction. When θ≥π, θ minus π; Solar irradiance I s The value range is [0,I s5max ], I s5max The maximum solar irradiance of the temperature tower measured in the past five years; Ambient temperature T A The value range is [T A5min ,T A5max ], T A5min and T A5max They are respectively the minimum and maximum values ​​of the tower ambient temperature measured in the past five years.

10. The method for dynamic capacity increase of power transmission lines based on online training of Bayesian optimization random forest according to claim 1, characterized in that: The weather forecast data are hourly forecast values ​​for the next 6 hours from the current hour, and the conductor current and conductor temperature of the transmission line are the measured average values ​​for the current hour.

Images