CN110807508B - Bus Peak Load Forecasting Method Considering Complicated Meteorological Effects - Google Patents

Abstract

A bus peak load prediction method considering complex weather influence belongs to the technical field of bus peak load prediction. According to the invention, the feature importance results of the condition mutual information on the features to be selected in the original feature set are used as a basis, the IPSO-ELM is combined as a predictor, forward feature selection is carried out, the optimal feature set of the bus peak load prediction is determined, the influence of feature redundancy on the prediction precision in the bus peak load prediction is reduced, the optimal prediction model is respectively constructed for different buses, the prediction precision of different buses is effectively improved, an improved particle swarm optimization extreme learning machine is introduced to be combined with a linear method, the peak load prediction under different scenes is carried out, and the prediction needs under small-sample or sample-free scenes are met.

Description

Bus Peak Load Forecasting Method Considering Complex Meteorological Effects

technical field

The invention belongs to the technical field of busbar peak load forecasting, and in particular relates to a busbar peak load forecasting method considering complex meteorological influences.

Background technique

The historical data of busbar peak load is limited, fluctuates violently and presents nonlinear and random characteristics, and the problem of low forecasting accuracy and difficult forecasting has become an urgent problem to be solved. How to improve the forecasting accuracy of busbar peak load has become an urgent problem to be solved. Bus peak load forecasting is an important basis for ensuring reliable and stable operation of power systems. It is of great significance to analyze and study methods to improve the accuracy of bus peak load forecasting.

At present, there have been many studies on the bus load forecasting, which optimized the bus load forecast according to the characteristics of the bus load, but did not fully analyze the influence of various factors such as natural weather and society on the bus load. When considering many factors, there is no characteristic selection, and did not take into account the different influencing factors of different busbars, did not carry out targeted feature selection for the influencing factors of different busbar loads, and did not establish a targeted busbar load forecasting model on this basis. Existing research on peak load forecasting is mostly carried out on the influencing factors and forecasting methods of peak load in city-level power grids. Although the existing research has improved the accuracy of peak load forecasting to a certain extent, there is no research on peak load forecasting with limited historical data. Carry out targeted analysis of sample questions.

Therefore, there is an urgent need for a novel technical solution in the prior art to solve this problem.

Contents of the invention

The technical problem to be solved by the present invention is to provide a bus peak load forecasting method that takes into account the influence of complex weather to solve the problem of small samples that have not yet been addressed in peak load forecasting, peak load fluctuations and limited historical data in the prior art Conduct targeted analysis of technical issues.

The bus peak load forecasting method considering complex meteorological influences includes the following steps, and the following steps are carried out in sequence,

Step 1. Based on the influence of natural meteorological and social factors on the accuracy of bus peak load forecasting, the original feature set of bus peak load forecasting is constructed according to natural meteorological and social factors, and the original feature set is analyzed by conditional mutual information (Conditional Mutual Information, CMI). Correlation between features and bus peak load, to obtain feature importance ranking;

Step 2. Optimize the input weight and threshold of the extreme learning machine by improving the particle swarm optimization algorithm to obtain the improved particle swarm optimization extreme learning machine (Improved Particle Swarm Optimization-Extreme Learning Machine, IPSO-ELM), and then take the prediction accuracy of IPSO-ELM as Decision variables, according to the feature importance ranking results obtained in step 1, carry out targeted forward feature selection to obtain the optimal feature subset for bus peak load forecasting, and retrain IPSO-ELM according to the obtained optimal feature subset to obtain the optimal Bus peak load forecasting model;

Step 3. Substituting the test set data within the set time period in the historical data into the optimal bus peak load prediction model obtained in step 2 to obtain the predicted value of the bus peak load;

Step 4. Through the linear model, the peak load of the bus under extreme high temperature conditions and the peak load of the bus under extremely low temperature conditions are respectively counted. According to the statistical results, it is known that the peak load of the bus has a linear relationship with the extreme high temperature and the extreme low temperature;

Step 5. Use the least square method to linearly fit the extreme high temperature and extreme low temperature and the peak load of the busbar, and obtain the linear models for peak load prediction of busbar 1 and busbar 2 under extreme temperature conditions respectively as follows:

Under extreme high temperature conditions, the linear model for busbar 1 peak load prediction is:

Under extreme low temperature conditions, the linear model of bus 1 peak load prediction is:

Under extreme high temperature conditions, the linear model for busbar 2 peak load prediction is:

Under extreme low temperature conditions, the linear model for bus 2 peak load prediction is:

为极端高温条件下母线1的峰值负荷；/>

为极端高温条件下母线2的峰值负荷；/>

为极端低温条件下母线1的峰值负荷；/>

为极端低温条件下母线2的峰值负荷；T为极端温度。formula:

is the peak load of bus 1 under extremely high temperature conditions; />

is the peak load of bus 2 under extremely high temperature conditions; />

is the peak load of bus 1 under extremely low temperature conditions; />

is the peak load of bus 2 under extremely low temperature conditions; T is the extreme temperature.

The natural weather includes longitude, latitude, temperature, air pressure, humidity, wind direction and wind speed.

The social factors include dates and holidays.

The prediction precision in described step 2 is the average absolute percentage error (MeanAbsolute Percent Error, MAPE) of predicted value and actual value, and the calculation formula of MAPE is:

为母线峰值负荷预测值；N为预测样本数。In the formula, Y _i is the measured value of bus peak load;

is the bus peak load forecast value; N is the number of forecast samples.

In said step 2, the specific method for obtaining IPSO-ELM by improving the particle swarm optimization algorithm to optimize the input weight and threshold of the extreme learning machine is as follows:

1) Construct the ELM model of bus peak load forecasting based on the original feature set, and randomly generate the input weight ω and hidden layer unit bias threshold b of ELM;

2) Determine the data of the original feature set required for input bus peak load prediction, and set the number of iterations;

3) Normalize the original feature set data for bus peak load prediction;

4) Obtain the fitness value according to the average absolute percentage error value of the bus peak load prediction of the ELM, and obtain the current individual fitness value and the group best fitness value;

5) On the basis of the particle swarm optimization algorithm PSO, according to the calculation formula of the nonlinear dynamic inertia weight coefficient and the calculation formula of the learning factor, the particle velocity and position are updated;

6) Calculate and obtain the current particle fitness value, compare the current particle fitness value with the historical optimal individual particle fitness value, if the current particle fitness value is better, then update the particle optimal solution; otherwise maintain the individual optimal fitness value ;

7) If the fitness value of the current particle is better than the swarm optimal solution, update the swarm optimal solution; otherwise, keep the swarm optimal solution unchanged;

8) If the number of iterations set in step 2) is not reached, return to step 3) until the set number of iterations is reached, and then substitute the optimal solution input weight ω and hidden layer unit bias threshold b into the ELM to construct the bus peak load predictive model.

The expression of the nonlinear dynamic inertia weight coefficient is:

In the formula: w is the inertia weight, w _min is the minimum value of the inertia weight, w _max is the maximum value of the inertia weight; f is the particle fitness; f _avg is the average fitness; f _min is the minimum fitness.

The expression of the learning factor is:

In the formula: c ₁ and c ₂ are learning factors, c _1s is the initial value of c ₁ , c _2s is the initial value of c ₂ , c _1c is the final value of c ₁ , c _2c is the final value of c ₂ ; iter is The number of current iterations; iter _max is the number of total iterations.

The number of iterations is 200.

Through the above design scheme, the present invention can bring the following beneficial effects:

The present invention is based on the conditional mutual information on the feature importance results of the features to be selected in the original feature set, and combines IPSO-ELM as a predictor to carry out forward feature selection to determine the optimal feature set for bus peak load prediction, reducing the bus load. The impact of feature redundancy on prediction accuracy during peak load forecasting, and the construction of optimal forecasting models for different buses can effectively improve the forecasting accuracy of different buses, and the combination of IPSO-ELM and linear methods is introduced to carry out peak load forecasting under different scenarios to meet Prediction needs in small-sample or no-sample scenarios.

Description of drawings

The present invention will be further described below in conjunction with accompanying drawing and specific embodiment:

Fig. 1 is a scatter diagram and a fitting curve diagram of the peak load average value of bus 1 under extremely high temperature conditions in the embodiment of the bus peak load forecasting method considering complex meteorological influences of the present invention.

Fig. 2 is a scatter diagram and a fitting curve diagram of the peak load average value of bus 2 under extremely high temperature conditions in the embodiment of the bus peak load forecasting method considering complex weather effects of the present invention.

Fig. 3 is a scatter diagram and a fitting curve diagram of the peak load average value of bus 1 under extremely low temperature conditions in the embodiment of the bus peak load forecasting method considering complex weather effects of the present invention.

Fig. 4 is a scatter diagram and a fitting curve diagram of the peak load average value of bus 2 under extremely low temperature conditions in the embodiment of the bus peak load forecasting method considering complex meteorological influences of the present invention.

Fig. 5 is an analysis diagram of the importance degree of peak load characteristics of bus 1 in the embodiment of the bus peak load prediction method considering complex meteorological influences in the present invention.

Fig. 6 is an analysis diagram of the importance degree of peak load characteristics of bus 2 in the embodiment of the bus peak load prediction method considering complex meteorological influences in the present invention.

Fig. 7 is an optimal feature set selection diagram of conditional mutual information combined with different predictors in an embodiment of the bus peak load forecasting method considering complex meteorological influences in the present invention.

Fig. 8 is a comparison chart of bus peak load forecasting between the targeted model and the unified model in the embodiment of the bus peak load forecasting method considering complex meteorological influences in the present invention.

Fig. 9 is a comparison chart of bus peak load forecasting of three methods in the embodiment of the method for forecasting bus peak load considering complex meteorological influences in the present invention.

Detailed ways

As shown in the figure, the bus peak load forecasting method considering complex meteorological influences includes the following steps:

1. Construction of original feature set for bus peak load forecasting

(1) The original feature set of bus peak load forecasting

Various factors such as natural weather and society will cause the peak load of the bus to fluctuate. Through the analysis of complex meteorological factors and related literature research, the original feature set constructed is shown in Table 1.

Table 1 The original feature set of bus peak load forecasting

Note:

1) _FT represents the peak temperature on the forecast day, F _Tave represents the average temperature on the forecast day; _FT(max,d-1) represents the peak temperature on the day before the forecast day, and _FT(ave,d-1) represents the day before the forecast day F _A is the air pressure at the forecast time, F _H is the humidity at the forecast time, F _W is the wind direction at the forecast time, F _W1 is the wind speed at the forecast time, F _A , F _H , F _W and F _W1 are all meteorological features;

2) F _G1 represents the longitude of the bus to be predicted, F _G2 represents the latitude of the bus to be predicted; F _T , F _Tave , F _T(max,d-1) , F _T(ave,d-1) , F _A , F _H , F _W , F _W1 , F _G1 and F _G2 are all natural meteorological features;

3) F _D1 to F _D7 are days of the week; F _J1 is a working day; F _J2 is a non-working day; F _H1 is a holiday; F _H2 is a normal day; F _D1 to F _D7 , F _J1 , F _J2 , F _{Both H1} and F _H2 are social factors;

4) FL _(max,d-1) represents the peak load of the bus one day before the forecast date, FL _(max,d-2) represents the peak load of the bus two days before the forecast date, and so on; _FL(t-15) FL _(t-30) represents the bus load 30 minutes before the peak load of the bus on the day before the test day, and so on.

(2) Conditional mutual information

In bus peak load forecasting, let D be the original feature set including various factors such as natural weather and society; Q be the measured bus peak load value; Z set is the selected feature. The mutual information between D and Q is:

In formula (1), F(D; Q) represents the mutual information between D and Q, P(d) is the marginal density function of D, P(q) is the marginal density function of Q, and P(d,q) is Joint probability density of D and Q.

Under the condition that the Z set is known, the conditional mutual information between the set D and the measured bus peak load value Q is:

In formula (2), F(D,Q|Z) represents the conditional mutual information between D and Q under Z condition, P(d|z) is the probability density function of D and Q under Z condition; P(q |z) is the probability density function of Q under Z condition; P(d,q|z) is the joint probability density function of D and Q under Z condition; P(d,q,z) is D, Q, Z The joint probability density function of .

There are many factors affecting bus peak load. If all the influencing factors are considered, information redundancy will be caused, and the prediction accuracy of bus peak load will be low. In order to improve the prediction accuracy of bus peak load, the correlation between various influencing factors and different bus peak loads is analyzed through conditional mutual information, and the ranking of feature importance in the original feature set is obtained.

2. Construction of busbar peak load forecasting model

(2.1) The principle of extreme learning machine based on improved particle swarm optimization

It is difficult to obtain an ideal bus peak load forecasting model by training predictors such as neural networks with limited bus peak load data. Therefore, an extreme learning machine suitable for small-sample training is applied to build a bus peak load predictor. In order to avoid the influence of improper parameter selection on the prediction effect, the improved particle swarm optimization algorithm is applied to optimize the input weight and threshold of the extreme learning machine, bus peak load predictor, to further improve the prediction accuracy.

(2.1.1) Extreme Learning Machine

其中，输入数据为x_i＝[x_i1,x_i2,…,x_in]^T∈Rⁿ，目标输出值为t_i＝[t_i1,t_i2,…,t_im]^T∈R^m。则隐层节点的数目为L的单隐层神经网络的ELM极限学习机网络模型可表示为With N samples

Wherein, the input data is _xi =[x _i1 , _xi2 ,…,x _in ] ^T ∈ R ⁿ , and the target output value is t _i =[t _i1 ,t _i2 ,…,t _im ] ^T ∈ R ^m . Then the ELM extreme learning machine network model of the single hidden layer neural network whose number of hidden layer nodes is L can be expressed as

In the formula, o _j represents the network output value; g represents the activation function; ω _i is the input weight; β _i is the output weight; b _i is the bias of the i-th hidden layer unit; x _j represents the data in _xi .

When there is no error, the activation function is infinitely close to any N samples, that is

According to formula (4), we can get

In the formula, g represents the activation function; ω _i is the input weight; β _i is the output weight; b _i is the bias of the i-th hidden layer unit; t _j represents the network output when the activation function can approach any N samples with zero error value;

The matrix form of N equations in formula (5) is

Hβ=T (6)

in,

H represents the hidden layer node output; T represents the desired output; β represents the weight matrix between the output layer and the hidden layer.

和/>

实现对单隐层ELM训练，使得by obtaining

and />

Realize the training of single hidden layer ELM, so that

表示β_i最优解；/>

表示ω_i的最优解；/>

表示b_i的最优解；H表示隐含层节点输出；T为期望输出；In the formula,

Indicates the optimal solution of β _i ; />

Indicates the optimal solution of ω _i ; />

Represents the optimal solution of _bi ; H represents the hidden layer node output; T is the expected output;

The minimum loss function equivalent to (9) is

In the formula, E represents the minimum loss function; g represents the activation function; ω _i is the input weight; β _i is the output weight; b _i is the bias of the ith hidden layer unit.

In ELM, once the input weight ω and the bias threshold b of the hidden layer unit are randomly determined, the unique hidden layer node output H can be obtained. From this, the ELM structure is determined.

(2.2) Parameter optimization of extreme learning machine based on improved particle swarm (2.2.1) Principle of improved particle swarm

In order to optimize the problem that the particle swarm algorithm is prone to fluctuations around the global optimal solution, the traditional particle swarm optimization is improved through the inertia weight, and its expression is:

In formula (11): w represents inertia weight, w _min is the minimum value of inertia weight, w _max is the maximum value of inertia weight; f is particle fitness; f _avg is average fitness; f _min is minimum fitness.

In order to enable the particle swarm optimization algorithm to quickly determine the global optimal solution, the traditional particle swarm is dynamically adjusted through formula (12), namely

In (12): c ₁ and c ₂ are learning factors, c _1s is the initial value of c ₁ , c _2s is the initial value of c ₂ , c _1c is the final value of c ₁ ; c _2c is the final value of c ₂ Value; iter is the number of current iterations; iter _max is the total number of iterations.

(2.2.2) ELM parameter optimization based on improved particle swarm

Taking the optimal process of constructing the bus peak load forecasting model with the original feature set as an example, the IPSO optimization ELM process is shown as follows:

1) Construct busbar peak load prediction ELM based on the original feature set, and randomly generate ELM input weight ω and hidden layer unit bias threshold b;

2) Determine the data of the original feature set required for input bus load peak prediction, and set the number of iterations;

3) Carry out normalization processing for the data of the original feature set of bus peak load prediction;

4) Obtain the fitness value according to the average absolute percentage error of the peak load prediction of the ELM, and determine the best fitness value of the current individual and group;

5) On the basis of traditional PSO, and according to equations (11) and (12), update particle velocity and position;

6) First, calculate the current particle fitness value, and then compare it with the historical optimal value. If it is better, update the particle optimal solution; otherwise, maintain the individual optimal fitness value;

7) If the fitness value of the current particle is better than the swarm optimal solution, update the swarm optimal solution; otherwise, keep the swarm optimal solution unchanged;

8) If the number of iterations is not reached, return to 3), otherwise, substitute the optimal solution ω and b into the ELM to construct a bus peak load forecasting model.

In the following feature selection link, the above methods are used to construct targeted optimal ELM predictors for feature sets of different dimensions.

3. No training sample bus peak load forecasting model

Although predictors such as ELM are suitable for small-sample forecasting, their construction still relies on historical samples. If there are extreme temperatures (extreme high temperature and extreme low temperature) that have not occurred in history on the forecast date, the forecast effect cannot be guaranteed. In order to improve the prediction accuracy of bus peak load without training samples, a linear model is introduced to predict the peak load of bus in extreme high and low temperature scenarios without training samples.

Example:

Statistical analysis was carried out on the peak loads of multiple buses under the extreme weather conditions of extreme high temperature and extreme low temperature in a certain city in Northeast China in 2018. In this embodiment, the extreme high temperature is set as the daily maximum temperature higher than 30 degrees, and the extreme low temperature is set as the daily minimum temperature lower than minus 20 degrees. According to the statistical results, there is an obvious linear relationship between the peak load of the busbar and the maximum temperature. Therefore, the linear fitting between the extreme temperature and the peak load of the busbar is carried out by using the least square method.

Figures 1 to 4 respectively show the scatter diagrams and corresponding fitting curves of the bus peak load average under the extreme high temperature and extreme low temperature weather conditions of bus 1 and bus 2 in a northeastern city in 2018. From Figures 1 to 4 and the linear model of the bus peak load prediction under the extreme weather conditions of the two buses, it can be seen that the impact of extreme weather on the peak load of different buses is different. Therefore, in extreme weather conditions, for different buses, targeted modeling analysis is required to improve the prediction accuracy of bus peak load.

The linear model of the peak load prediction of bus 1 and bus 2 under extreme meteorological conditions obtained by using the least square method to linearly fit the extreme temperature and bus peak load is as follows:

Under extreme high temperature conditions, the linear model for busbar 1 peak load prediction is:

Under extreme low temperature conditions, the linear model of bus 1 peak load prediction is:

Using the same method, get the bus 2 linear prediction model.

Under extreme high temperature conditions, the linear model for busbar 2 peak load prediction is:

Under extreme low temperature conditions, the linear model for bus 2 peak load prediction is:

和/>

分别为极端高温条件下母线1和母线2的峰值负荷；/>

和/>

分别为极端低温条件下母线1和母线2的峰值负荷；T为极端温度。formula:

and />

are the peak loads of busbar 1 and busbar 2 under extreme high temperature conditions respectively; />

and />

are the peak loads of busbar 1 and busbar 2 under extremely low temperature conditions; T is the extreme temperature.

In the study, the busbar peak load and meteorological information data of a city in Northeast China in 2018 were used. The data in January, April, August, and October were used as the verification set, the data in July was used as the test set, and the remaining 7 months were used as the training set. Sexual selection. In order to prove the advanced nature of the new method, IPSO-ELM, ELM and BP neural network (Back Propagation NeuralModel, BPNN) were used as comparative experiments. The prediction effect of the model is evaluated by mean absolute percentage error (Mean Absolute Percent Error, MAPE) and root mean square error (Root Mean Square Error, RMSE). The index calculation method is as follows:

为母线峰值负荷预测值；N为预测样本数。In formulas (17) and (18): Y _i is the measured value of bus peak load;

is the bus peak load forecast value; N is the number of forecast samples.

(1) Feature selection analysis based on conditional mutual information

Figure 5 and Figure 6 are the analysis diagrams of the importance of peak load characteristics of different buses. It can be seen from the figure that there are differences in the correlation between peak load and characteristics of different buses. Figure 7 shows the number of features contained in the optimal feature set and the corresponding MAPE values after CMI combines IPSO-ELM, BPNN, and ELM to select the optimal feature set for bus 1 respectively; Table 2 shows the features included in the optimal feature set . It can be seen from Figure 7 that when the bus peak load is predicted, the selected feature sets are different, and the prediction errors are different. When IPSO-ELM, BPNN, and ELM are used as predictors respectively, and the feature dimensions of feature subsets are 25, 28, and 34, respectively, the error of bus peak load prediction is the smallest. And it can be seen from Figure 7 that, according to the optimal feature set, when BPNN, ELM, and IPSO-ELM are used as predictors, the MAPE values of bus 1 peak load prediction are 4.54%, 3.75%, and 3.04%, respectively. Among the sensors, the MAPE value of IPSO-ELM bus peak load prediction is the smallest, which shows the advantages of high prediction accuracy of IPSO-ELM. Similarly, the optimal feature subset of bus 2 can be determined.

Table 2 Optimal feature set

(2) Analysis of busbar peak load forecast results

In order to prove the advantages of the proposed bus peak load forecasting method with high accuracy, the peak load forecast results of bus 1 and bus 2 in a city in Northeast China in July 2018 are listed. Figure 8 shows the peak load prediction results of the two buses based on targeted modeling and the prediction model constructed using the original feature set; Table 3 corresponds to the peak load prediction error of the bus in Figure 8. Under the optimal model, the MAPEs of the peak load predictions of the two buses are 3.04% and 2.98% respectively; when the original feature set is modeled, the MAPEs of the peak load predictions of the two buses are 3.89% and 4.01%, respectively. It can be seen from the comparison that the targeted feature selection is carried out for different buses, and the bus peak load prediction model is established according to the feature selection results, with high prediction accuracy.

Figure 9 shows the bus peak load forecast results of bus 1 in a certain city in Northeast China in July 2018 when IPSO-ELM, ELM and BPNN are used as bus peak load predictors respectively. Table 4 corresponds to the peak load forecast of the three buses in Figure 9 The prediction error of the method. It can be seen from Table 4 that the MAPE of bus peak load prediction using BPNN, ELM and IPSO-ELM is 4.23%, 3.93%, and 3.04%, respectively, and the MAPE of IPSO-ELM bus peak load prediction is the smallest. Therefore, the method proposed by the present invention has high prediction accuracy for the problem that complex meteorological factors cause severe fluctuations in the peak load of the bus.

Table 3 Bus Peak Load Prediction Error

Table 4 Peak load forecasting errors of the three methods

In order to further verify the effectiveness of the new method in extreme high temperature and extreme low temperature scenarios without training samples, the highest and lowest temperature days in a certain city in Northeast China in 2018 were removed from the training samples, and the highest and lowest temperature days were used as the days to be predicted. The historical data before the temperature day is used as training data. The improved particle swarm optimization extreme learning machine and linear model were used to predict the peak load of different buses on that day, and the results are shown in Table 5. It can be seen from Table 5 that the linear model has higher accuracy in predicting the bus peak load of samples without historical temperature, which proves that the new method can effectively avoid the peak load prediction error caused by the lack of historical data under extreme temperatures without training samples. The new method has better applicability.

Table 5 Bus Peak Load Forecast on Extreme Weather Days

In summary, based on the conditional mutual information value, the feature selection of bus peak load forecasting is carried out, which reduces the influence of feature redundancy on the prediction accuracy of bus peak load forecasting, and the optimal forecasting models for different buses are constructed to effectively improve the performance of different busbars. In order to improve the prediction accuracy, the combination of improved particle swarm optimization extreme learning machine and linear method is introduced to carry out peak load prediction in different scenarios to meet the prediction needs in small sample or no sample scenarios.

Claims (5)

1. A bus peak load prediction method considering complex weather influence is characterized in that: comprising the following steps, and the following steps are carried out in sequence,

firstly, constructing an original characteristic set for predicting the bus peak load according to natural weather and social factors based on the influence of the natural weather and the social factors on the bus peak load prediction precision, and analyzing the correlation between each characteristic in the original characteristic set and the bus peak load through conditional mutual information (Conditional Mutual Information, CMI) to obtain characteristic importance ranking;

step two, optimizing the input weight and the threshold value of the extreme learning machine through an improved particle swarm algorithm to obtain an improved particle swarm optimized extreme learning machine (Improved Particle Swarm Optimization-Extreme Learning Machine, IPSO-ELM), then carrying out targeted forward feature selection according to the feature importance ranking result obtained in the step one by taking the IPSO-ELM prediction precision as a decision variable to obtain an optimal feature subset for bus peak load prediction, and retraining the IPSO-ELM according to the obtained optimal feature subset to obtain an optimal bus peak load prediction model;

the specific method for obtaining the IPSO-ELM by optimizing the input weight and the threshold value of the extreme learning machine through the improved particle swarm algorithm comprises the following steps:

1) Constructing an ELM model for predicting bus peak load based on the original feature set, and randomly generating an input weight omega of the ELM and a hidden layer unit bias threshold b;

2) Determining data of an original feature set required by input bus peak load prediction, and setting iteration times;

3) Predicting original feature set data aiming at bus peak load, and carrying out normalization processing;

4) Obtaining a fitness value according to the average absolute percentage error value of the bus peak load prediction of the ELM, and obtaining a current individual fitness value and a group optimal fitness value;

5) Based on a particle swarm algorithm PSO, updating the particle speed and the position according to a calculation formula of a nonlinear dynamic inertia weight coefficient and a learning factor calculation formula;

6) Calculating and obtaining a current particle fitness value, comparing the current particle fitness value with a historical optimal individual particle fitness value, and updating a particle optimal solution if the current particle fitness value is better; otherwise, maintaining the optimal fitness value of the individual;

7) The fitness value of the current particle is better than the optimal solution of the group, and the optimal solution of the group is updated; otherwise, maintaining the optimal solution of the population unchanged;

8) Returning to the step 3) until the set iteration times are not reached, substituting the optimal solution input weight omega and the hidden layer unit bias threshold b into ELM, and constructing a bus peak load prediction model;

the expression of the nonlinear dynamic inertia weight coefficient is as follows:

wherein: _w is the inertial weight, w _min Is the minimum value of inertia weight, w _max Is the maximum value of the inertial weight; f is the particle fitness; f (f) _avg Is the average fitness; f (f) _min Is the minimum fitness;

the expression of the learning factor is:

wherein: c ₁ And c ₂ C is a learning factor _1s C is ₁ Initial value of c _2s C is ₂ Initial value of c _1c C is ₁ Final value of c _2c C is ₂ A final value of (2); item is the number of current iterations; ter (iter) _max Is the total number of iterations;

substituting the test set data in the set time period in the historical data into the optimal bus peak load prediction model obtained in the step two to obtain a predicted value of the bus peak load;

step four, respectively counting the bus peak load under the extreme high temperature condition and the bus peak load under the extreme low temperature condition through a linear model, and obtaining that the bus peak load and the extreme high temperature and the extreme low temperature have linear relations according to the counting result;

step five, carrying out linear fitting on the extreme high temperature and the extreme low temperature and the bus peak load by using a least square method, and respectively obtaining the linear models of the peak load prediction of the bus 1 and the bus 2 under the extreme temperature condition as follows:

under the extreme high temperature condition, the linear model of the bus 1 peak load prediction is as follows:

under the extremely low temperature condition, the linear model of the bus 1 peak load prediction is as follows:

under the extreme high temperature condition, the linear model of the bus 2 peak load prediction is as follows:

under the extremely low temperature condition, the linear model of the bus 2 peak load prediction is as follows:

in the formula:

peak load of the bus bar 1 under extremely high temperature conditions; />

Peak load of the bus bar 2 under extremely high temperature conditions; />

Peak load of bus bar 1 at extremely low temperature; />

Peak load of the bus bar 2 under extremely low temperature conditions; t is the extreme temperature.

2. The bus peak load prediction method considering complex weather effects according to claim 1, wherein: the natural weather includes longitude, latitude, temperature, air pressure, humidity, wind direction and wind speed.

3. The bus peak load prediction method considering complex weather effects according to claim 1, wherein: the social factors include date and holiday.

4. The bus peak load prediction method considering complex weather effects according to claim 1, wherein: the prediction precision in the second step is the average absolute percentage error (Mean Absolute Percent Error, MAPE) between the predicted value and the actual value, and the calculation formula of MAPE is as follows:

in the formula, Y _i The actual measurement value of the bus peak load;

predicting a bus peak load value; n is the number of predicted samples.

5. The bus peak load prediction method considering complex weather effects according to claim 1, wherein: and the iteration times in the second step are 200 times.

Images