CN-121980947-A - Method for establishing diversified electricity consumption prediction model for micro-grid unit

CN121980947ACN 121980947 ACN121980947 ACN 121980947ACN-121980947-A

Abstract

The invention belongs to the technical field of electric power, and discloses a method for establishing a diversified electricity utilization prediction model for a micro-grid unit, which is characterized by comprising the following steps of: the method comprises the steps of constructing a power distribution system source network charge storage layered partition structure, constructing a diversified electricity demand model for a micro-grid unit, predicting distributed photovoltaic output based on space-time fusion, predicting charging load of an electric automobile and predicting response load of a demand side. The method has the main beneficial effects that the prediction accuracy under the condition of uncertainty of new energy sources is improved, the daily short-term prediction accuracy is more than 90%, an aggregation calculation method of the charging load of the electric automobile is established, a stripping method and a prediction model based on a daily reference load curve are provided, the correlation between the stripping method and the temperature is analyzed, and the daily short-term load prediction accuracy is more than 95%.

Inventors

SHEN YU
CHEN WEI
CHEN CHAO
XIONG JIE
LIU WEI
SONG ZIJIAN
YU LIANGMING
CHENG LONG
SHAO JIE
XIONG XIAOKUN
YANG SHIJIE
ZHANG WEILIN
SHEN RUO
LIU XUANCHENG
DAI XIAOCONG

Assignees

国网湖北省电力有限公司鄂州供电公司

Dates

Publication Date: 20260505
Application Date: 20260127

Claims (10)

1. The establishment method of the diversified electricity consumption prediction model for the micro-grid unit is characterized by comprising the following steps of: constructing a power distribution system source network charge storage layered partition structure; Constructing a diversified electricity demand model for the micro-grid unit; A step of predicting distributed photovoltaic output based on space-time fusion; Predicting the charging load of the electric automobile; And a step of demand side response load prediction.
2. The method for establishing the diversified electricity consumption prediction model for the micro-grid unit according to claim 1 is characterized in that the step of predicting the distributed photovoltaic output based on space-time fusion comprises the steps of constructing prediction thought and used model principle, and predicting the output of a photovoltaic power station and predicting interval probability.
3. The method for establishing the diversified electricity consumption prediction model for the micro-grid unit according to claim 1 is characterized in that the step of predicting the charging load of the electric automobile comprises the steps of constructing a charging strategy of the electric automobile, predicting the remaining amount of the electric automobile and calculating the charging load of the whole electric automobile.
4. The method for building the diversified electricity consumption prediction model for the micro-grid unit according to claim 1, wherein the step of predicting the demand side response load comprises the steps of cooling load, load calculation and prediction.
5. The method for building the diversified electricity consumption prediction model for the micro grid unit according to claim 2, wherein in the step of building the prediction idea and the used model principle, the main two core contents of the K-shape algorithm are a distance measurement method based on cross correlation and calculation of a time sequence clustering center, wherein in the distance measurement based on cross correlation, the scale change is that the time sequence has a difference in scale, and the sequence Conversion into Wherein a and b are constants, the similarity between X and Y is unchanged, the displacement is changed, the phase of the two time sequences has a certain deviation, The similarity between X and Y is unchanged, the noise is changed when two time sequences have similar forms but are interfered by noise of different degrees, and the cross correlation is used for comparing two sequences And And (3) moderately translating a sample X time window by utilizing a mutual correlation relation of the two sequences, and comparing global shape characteristics of the two sequences: ; In the middle of Is that A time series after the time window is translated; Is that All possible amounts of translation of the time window, wherein If (1) Then Is shifted to the right Units of, if Then Is shifted to the left Units of units from which cross-correlation sequences are derived Having a length of The definition is as follows: And wherein: Calculating the time When reaching the maximum value Relative to the position of (2) , Is the optimal displacement of Cross-correlating the sequence coefficients Normalized to The following formula is obtained so that the fluctuation range is between-1 and 1, wherein the larger the coefficient value of the cross-correlation sequence is, the higher the positive correlation degree of two sequences is: Wherein: obtaining the correlation coefficient value corresponding to the completely similar time sequence when the time sequence does not have relative displacement, obtaining the time sequence similarity judgment The separation measurement method is characterized in that the calculation formula of the separation measurement method after per unit is as follows: Wherein: The values range from 0 to 2, where 0 represents that the time series samples are completely similar, using The distance is used for carrying out similarity measurement on the time sequence and is used as the basis for the subsequent cluster classification.
6. The method for building the diversified electricity consumption prediction model for the micro-grid unit according to claim 5, wherein the time sequence clustering center calculating method is that the clustering center is extracted to be regarded as an optimization problem, and the aim is to find a sequence with the minimum sum of squares of each type of time sequences, namely, a Steiner tree optimization problem: Wherein: Is the first A cluster-like data set; In each iteration, all sequences are aligned with the reference sequence by using the previous iteration center as a reference and a cross-correlation method: after normalizing it: Wherein: , wherein For a matrix with diagonal elements of 1 and other elements of 0, Is a matrix of all 1's, By using Instead of ; Maximum value is matrix The feature vector corresponding to the maximum feature value is the extracted typical load characteristic curve and is also the clustering center of each type.
7. The method for building a multi-element power consumption prediction model for a micro-grid unit as set forth in claim 6, wherein the K-shape algorithm comprises the following specific procedures of using time sequence data by the K-shape algorithm And the number of clusters For input, a cross-correlation method is utilized to find out the center of the class cluster, and the sample data of each class cluster is continuously updated in each iteration, and the specific steps are as follows: Step 1, designating the clustering number Initializing each type of cluster center Wherein Is zero vector; Step 2, sequentially calculating a data set by using the per unit SBD distance calculation method Each of the loads of (a) To various centers Distance of (2) And will Is classified into sum Class with smallest distance In (a) and (b); Step 3, extracting each type of clustering morphological characteristics, namely each type of clustering centers; And 4, repeating the step 2 and the step 3. When no more changes occur in the cluster centers of each class, the iteration is stopped.
8. The method for establishing the diversified electricity consumption prediction model for the micro-grid unit according to claim 2, wherein in the step of establishing the prediction thought and the used model principle, the random forest algorithm flow is as follows: The number of the total training samples is N, and N training samples serving as a single decision tree are randomly extracted from N training sets by a single decision tree; The second step, when splitting is carried out on each node of each decision tree, M input features are randomly selected from the M input features, then the M input features are selected to be the best for splitting, M is not changed in the process of constructing the decision tree, M features are randomly selected for each node, then the best features are selected to be split, and two selection measures of splitting attributes in the decision tree, namely information gain and a base index, can be selected as one measurement method; Thirdly, each tree is split all the time until all training samples of the node belong to the same class, pruning is not needed, and the randomness is ensured by the previous two random sampling processes, so that the fitting phenomenon can not occur under the condition of not pruning; Photovoltaic power station output prediction and interval probability prediction, wherein photovoltaic power station data adopts sample data of certain photovoltaic power station power and influencing factors, and the photovoltaic output related data is expressed as The data were z-score normalized prior to analysis, as shown in the following formula: Wherein As an arithmetic mean of the data, Is the standard deviation of the data.
9. The method for building a diversified electricity consumption prediction model for a micro grid unit according to claim 3, wherein the charging load calculation step is as follows: step 1, probability statistical distribution data of an electric automobile travel rule and battery parameters are input; step 2, presetting the number of vehicles to be ; Step 3, a vehicle is driven in, ; Step 4, randomly sampling each electric automobile by using a Monte Carlo simulation method, and taking the electric automobile as the data of the in-out time and the initial charge state of the electric automobile; Step 5, according to the formula Calculating the charge time length and judging the charge state of each period, wherein And The measured wind speed of the monitoring point and the wind speed of the hub are respectively measured, the H and the Href are respectively the height of the hub and the height of the real measuring point, For the surface roughness description factor, The value of (2) is related to the ground surface environment of the field where the wind turbine generator is erected, and the value is 1/7-1/4; accumulating the charging load of the electric automobile to obtain the total charging load of the electric automobile according to the following formula; wherein For inverter conversion efficiency, the calculation formula of the constant coefficients a and b is as follows: In which, in the process, 、 Is the rated wind speed and rated power of the wind turbine; 、 The wind turbine is cut-in wind speed and cut-out wind speed, and when the cut-in wind speed is smaller than the minimum starting wind speed or larger than the cut-out wind speed, the wind turbine generator needs to be cut off for working; Step 7, judging whether the number of the preset electric vehicles is reached, if so, performing step 8, and if not, performing step 3; and 8, outputting a charging load curve of the electric automobile, and dividing the total charging load by the total number of the vehicles to obtain an average charging power curve of the single vehicle.
10. The method for building a power consumption prediction model for a micro-grid unit according to claim 4, wherein the cooling load is obtained by obtaining the maximum daily cooling load and the length is the maximum daily cooling load based on the daily reference load prediction A kind of electronic device Sequence and sequence of the sequence Correlation coefficient of sequence The calculation is as follows: , wherein, 、 Respectively is A sequence of, When the correlation coefficient between the daily maximum cooling load and the daily maximum temperature is more than 0.7, the correlation is strong, the variation between the daily maximum cooling load and the daily maximum temperature is more similar to the linear correlation, and a regression prediction model is adopted for predicting the daily maximum cooling load; The prediction is carried out by a GM (1.1) model, and the prediction method is as follows: first, assume an original data sequence: In which, in the process, To make the model easier to process these data, the original data is accumulated once to obtain another new sequence: In this new sequence, the sequence of the sequence, The cumulative value from the first data point to the t data point is represented, and the cumulative data sequence is described by a first order linear differential equation: In this equation, a is a parameter to be estimated called a development coefficient, u is another parameter called a gray action amount, which represents that a and u can be solved by regression analysis or least square method in addition to the growth trend, and the above differential equation is integrated to obtain the following expression: calculating an accumulated data value at any one time point t, wherein Is the initial accumulated data value that is to be accumulated, Representing the various points in time, by which future data trends can be predicted.

Description

Method for establishing diversified electricity consumption prediction model for micro-grid unit Technical Field The invention belongs to the technical field of power, and particularly relates to a method for establishing a diversified power consumption prediction model for a micro-grid unit. Background Under the guidance of a double-carbon strategic target, the construction of a novel power system mainly based on new energy becomes a necessary choice for energy transformation in China. In the novel power system, the coupling relation among the main network, the distribution network and the micro network is increasingly compact, and the operation scene is increasingly complex. The initiative of each level of power grid is obviously enhanced, and especially, as a large amount of distributed power sources, energy storage, flexible loads and the like are accessed into the distribution network and the micro-grid, the scale and controllable resources of the distribution network and the micro-grid are rapidly increased, the power grid structure tends to be equivalent and interactive, the influence of the lower level of power grid on the running state of the upper level of power grid is more obvious, and unprecedented higher requirements are put forward on the regulation running level. The traditional main-distribution-micro network cooperative mode mainly comprising static, unidirectional and rigid connection has difficulty in adapting to complex situations of multisource, multi-load and multi-coexistence under high-proportion new energy access. Therefore, aiming at the construction background of a novel power system, the key nodes of layered and partitioned power balance in the power distribution network are divided by taking the interaction of source network and load storage as a core target, the relationship of the key nodes is fully understood and built from the aspects of overall load, friendly load and distributed power output, the load diversity form in the power distribution network is researched, and the technical problems of predicting the diversity load, building a distributed power output prediction model and the like are urgently needed to be solved. CN111985701A discloses a power consumption prediction method based on a power supply enterprise big data model library, which comprises the following steps of S1, acquiring training data to train a power consumption prediction model in a power consumption prediction model library, and realizing updating of a model version of the power consumption prediction model library, S3, carrying out power consumption prediction by using the trained power consumption prediction model, wherein the power consumption prediction model library comprises a data preprocessing model, a short-term load prediction model and a space load prediction model, the data preprocessing model comprises a data cleaning module, a data standardization module and a data noise reduction module, and the short-term load prediction model comprises a basic prediction model, a support vector machine prediction model, an LTSM neural network prediction model and a platform cluster load prediction model, and the space load prediction model is used for predicting the size and the position of a future power load in a power supply area. The prediction accuracy under the condition of uncertainty of new energy, the prediction of the charging load of the electric automobile, the correlation between the daily reference load and the temperature and the like are required to be further improved. Disclosure of Invention In order to solve the problems, the invention aims to disclose a method for establishing a diversified electricity utilization prediction model for a micro-grid unit, which is realized by adopting the following technical scheme. The establishment method of the diversified electricity consumption prediction model for the micro-grid unit is characterized by comprising the following steps of: constructing a power distribution system source network charge storage layered partition structure; Constructing a diversified electricity demand model for the micro-grid unit; A step of predicting distributed photovoltaic output based on space-time fusion; Predicting the charging load of the electric automobile; And a step of demand side response load prediction. The method for establishing the diversified electricity consumption prediction model for the micro-grid unit is characterized in that the step of predicting the distributed photovoltaic output based on space-time fusion comprises the steps of constructing prediction thought and used model principle, and the steps of predicting the output of a photovoltaic power station and predicting interval probability. The method for establishing the diversified electricity consumption prediction model for the micro-grid unit is characterized in that the step of predicting the charging load of the electric automobile comprises the steps of constructing a chargin