A data fitting method for electricity load curve based on minimum interval dynamic distribution

Abstract

The present invention belongs to the field of application of power consumption data analysis and processing. In order to solve the problem of recovering abnormal power consumption indication curve data, a method for fitting power consumption load curve data based on minimum interval dynamic distribution is provided, including the following steps: S1: Identification and classification of user power consumption curve characteristics; S2: Construction of minimum interval dynamic characteristic curves of different user power consumption behaviors; S3: Similarity matching between daily power consumption behavior characteristic curves and minimum interval dynamic characteristic curves; S4: Data fitting based on power consumption weights with minimum power consumption intervals. The present invention is used to support grid enterprises in carrying out business such as load forecasting, trade settlement, electricity price formulation, and agency power purchase in the electricity market, as well as the comprehensive energy efficiency management of power users and the needs of major government analysis and decision-making; this algorithm can effectively ensure the data integrity of the user's power consumption indication curve and the accuracy of the fitted data, and overall improve the quality of the power consumption indication curve data.

Description

Technical Field

[0001] The present invention belongs to the technical field of power consumption data analysis and processing applications, and particularly relates to a method for fitting power load curve data based on minimum interval dynamic distribution. Background Art

[0002] To support the reform of the power market, the formulation of time-of-use electricity prices, and trade settlement, etc., State Grid Corporation relies on "State Grid Online" to launch the market-oriented power sales e-assistant, which opens the function of querying the historical electricity consumption load curves of users to market-oriented users and power sales companies, and issues the "Notice on Doing a Good Job in the Promotion of 'Market-oriented Power Sales e-Assistant' and the Electricity Bill Settlement Work of Power Generation Enterprises" (Marketing Business

[2021] No. 23), which also puts forward new requirements for accelerating the improvement of high-frequency meter reading capabilities and improving the quality of minute-level electricity curve data. Glitch data caused by accidental failures of equipment and communication error codes, and intermittent acquisition failures caused by reasons such as unstable remote channels and marketing meter replacement services, etc., make data loss inevitable, which will inevitably result in incomplete data of the electricity indication transaction curve and affect the accurate implementation of work such as user load forecasting, electricity price formulation, and proxy power purchase.

[0003] Most of the existing curve data prediction technologies in the power industry are aimed at the overall power consumption of the power grid, and its characteristics are relatively stable as a whole. However, the electricity indication curve of a single user is greatly affected by factors such as individual production behaviors and external weather changes, which will cause uncertainty and strong non-linearity to the electricity consumption trend. The existing data prediction and fitting methods only have reference value and cannot well meet the requirements of fitting the user transaction curve. How to fully combine the individual electricity consumption behavior characteristics of users to achieve the restoration and prediction of electricity curve data is the key problem to be solved.

[0004] Chinese Patent Application CN202110212361.2 discloses a method for predicting short-term daily load curves based on DCAE-LSTM, including: Step 1), constructing a dataset of power loads with high noise; Step 2), constructing a convolutional denoising autoencoder; Step 3), constructing a load prediction network based on a long short-term memory network; Step 4), combining the convolutional denoising autoencoder with the load prediction network based on the long short-term memory network to form a hybrid deep neural network; Step 5), inputting the dataset with high noise in Step 1) into the hybrid deep neural network in Step 4) to achieve the prediction of short-term daily load curves. Its deficiency is that although this method effectively eliminates the influence of high noise on load power consumption, it does not make full use of the individual behavior factors of users, and there are obvious shortcomings in the prediction of individual user load curves. Summary of the Invention

[0005] To solve the problems existing in the prior art, the present invention proposes a method for fitting power consumption load curve data based on minimum interval dynamic distribution, aiming to effectively recover and predict the power indication curve data of missing data points, restore the true situation of individual user power consumption curves, and improve the reliability of transaction settlement data.

[0006] The present invention proposes a method for fitting power consumption load curve data based on minimum interval dynamic distribution, comprising the following steps:

[0007] S1: Identification and classification of user power consumption curve characteristics;

[0008] S2: Constructing minimum interval dynamic characteristic curves for different user power consumption behaviors;

[0009] S3: Similarity matching between daily power consumption curves and minimum interval dynamic characteristic curves;

[0010] S4: Data fitting based on power consumption weights with minimum power consumption intervals.

[0011] The identification and classification of user power consumption curve characteristics described in step S1 are specifically implemented according to the following steps:

[0012] Take the power indication curves of power consumption customers in the recent D days, perform feature identification on each daily curve, and record the feature information of each daily curve; the features are described and recorded by identifying the time sequence of data points of sudden increases and decreases in interval power consumption and the number of sudden increases and decreases; the feature quantity recording sequence: Q = {M, N, {T1, T2, T3,..., T M+N}}, where M is the number of sudden increases, N is the number of sudden decreases, and is the xth mutation information element, including the mutation time point and mutation description;

[0013] Construct a feature information set of curves in the recent D days:

[0014]

[0015] The construction of minimum interval dynamic characteristic curves for different user power consumption behaviors described in step S2 is specifically implemented according to the following steps:

[0016] Merge similar power consumption characteristics according to curve feature parameters to form the power consumption feature classification of this user: {Q1, Q2, Q3,..., Q x}; Take all the curve data under a certain feature classification Q x and obtain the power consumption of each effective interval of each curve. Calculate the weighted average of the interval power consumption corresponding to the same time point to generate interval power consumption {q1, q2, q3,..., q 95}.

[0017] The similarity matching between the daily power consumption curve and the minimum interval dynamic characteristic curve described in step S3 is specifically implemented according to the following steps:

[0018] Obtain the current power consumption indication transaction curve data of the user;

[0019] According to the missing data points of the curve data, divide a curve data into multiple continuous curve segments, obtain the interval power consumption sequence corresponding to each curve segment, and form multiple power consumption curve segments;

[0020] According to the effective power consumption interval of the power consumption curve segment, use the Pearson coefficient method to calculate the correlation within the corresponding time series of this curve segment and each dynamic characteristic curve of the user;

[0021] Based on the correlation between each segmented curve of this curve and each dynamic characteristic curve, use the similarity weighted ratio method to calculate the similarity between the whole curve and each characteristic curve.

[0022] The calculation description of the curve segmentation correlation is as follows:

[0023] Suppose the curve segment q = [q1, q,..., q 14 , obtain the data y = [y1, y2,..., y 14 of the same period of a certain characteristic curve of this user

[0024] Then the calculation formula is as follows:

[0025]

[0026] Where ρ q，y is the required correlation coefficient, cov(q, y) is the covariance of the q and y vectors, and δ q , δ y are the standard deviations of the q and y vectors respectively. The covariance calculation formula is as follows:

[0027] cov(q, y) = E((q - μ q )(y - μ y ))

[0028] Where E represents expectation, and μ q , μ y represent the means of q and y respectively.

[0029] The data fitting based on the power consumption weight of the minimum power consumption interval described in step S4 is specifically implemented according to the following steps:

[0030] Obtain the two credible values before and after the missing data curve segment. According to the trend and total power consumption of the corresponding characteristic curve in the same period, use the weighted power ratio to obtain the fitted power of each interval Then obtain the curve fitting indication value according to the credible points; where, Q: the interval power consumption to be obtained; Q总 : Total power consumption of the curve with missing data points Power consumption corresponding to the required interval of the interval power consumption reference curve Total power consumption corresponding to the missing data points of the required curve of the interval power consumption reference curve

[0031] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0032] 1. To solve the problem of missing data points in the user's electricity consumption indication curve, a method for fitting electricity load curve data based on minimum interval dynamic distribution is proposed. This method first constructs a mathematical model of the electricity curve characteristics based on historical data, then classifies the user's electricity consumption behavior reliably, and then uses the same type of electricity consumption characteristics to effectively fit and predict the curve with missing data, completing the repair of the curve data. This method has high universality and accuracy in solving the problem of missing electricity curve data of individual users.

[0033] 2. The present invention provides a method for fitting electricity load curve data based on minimum interval dynamic distribution to support the needs of power grid enterprises in carrying out power market transaction load forecasting, trade settlement, electricity price formulation, agent power purchase and other services, comprehensive energy efficiency management of power users, and major analysis and decision-making of the government. This algorithm can effectively ensure the data integrity of the electricity indication curve and the accuracy of the fitted data, and overall improve the quality of the electricity indication curve data. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The above and / or additional aspects and advantages of the present invention will become apparent and be readily understood from the following description of the embodiments in conjunction with the accompanying drawings, wherein:

[0035] Figure 1 is the block diagram of the method of the present invention.

[0036] Figure 2 is the general flowchart of the process of the present invention.

[0037] Figure 3 is the flowchart for constructing the dynamic characteristic curve of the present invention.

[0038] Figure 4 is the flowchart for matching the best dynamic characteristic curve of similarity of the present invention.

[0039] Figure 5 is the flowchart for fitting the interval electricity consumption trend of the present invention.

[0040] Figure 6 is the schematic diagram of the user's electricity consumption characteristic curve of the present invention.

[0041] Figure 7 is the result diagram of the example simulation verification of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0042] Embodiments of the present invention will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the drawings are exemplary and are only used to explain the present invention and should not be construed as limiting the present invention.

[0043] As Figures 1-7 shown, the present invention is implemented as a method for fitting power consumption load curve data based on minimum interval dynamic distribution, including the following steps:

[0044] S1: Identification and classification of user power consumption curve characteristics;

[0045] S2: Constructing a minimum interval dynamic characteristic curve for different user power consumption behaviors;

[0046] S3: Similarity matching between the daily power consumption curve and the minimum interval dynamic characteristic curve;

[0047] S4: Data fitting based on the power consumption weight with the minimum power consumption interval.

[0048] The following is a detailed explanation of the specific steps involved in the above method:

[0049] Step 1, the identification and classification of user power consumption curve characteristics described in step S1 are specifically implemented according to the following steps:

[0050] Take the power consumption indication curves of electricity customers in the recent D days, perform feature identification on each daily curve, and record the feature information of each daily curve.

[0051] The curve characteristics defined in the present invention mainly consider two dimensions: time series and increase rate. For each daily power consumption indication curve, a value at a moment is recorded every 15 minutes from 00:00 to 23:45, with a total of 96 data points. This feature is described and recorded by identifying the time sequence of data points with sudden increases or decreases in interval power consumption and the number of sudden increases or decreases.

[0052] Feature quantity recording sequence: Q = {M, N, {T1, T2, T3,..., T M+N}}, where M is the number of sudden increases, N is the number of sudden decreases, and T x is the xth mutation information element, including the mutation time point and the mutation description (sudden increase or sudden decrease);

[0053] Construct a curve feature information set for the recent D days:

[0054]

[0055] Feature parameter identification:

[0056] 1) Identification of sudden increase in electricity consumption at intervals: Calculate the electricity consumption at each effective interval of the curve data in sequence. When the electricity consumption in two adjacent intervals, the latter is greater than the minimum scale and greater than K1 (K1 is a constant coefficient in the range of [2, +∞], and the value in this invention is 2) times of the former, the latter is considered as the interval with sudden increase in electricity consumption, and record the information of this sudden increase interval according to the definition requirements of element T. x Define the requirements and record the information of this sudden increase interval.

[0057] 2) Identification of sudden decrease in electricity consumption at intervals: Calculate the electricity consumption at each effective interval of the curve data in sequence. When the electricity consumption in two adjacent intervals, the latter is not zero and less than K2 (K2 is a constant coefficient less than 1, and the value in this invention is 0.5) times of the former, the latter is considered as the interval with sudden decrease in electricity consumption. And record the information of this sudden decrease interval according to the definition requirements of element T. x Define the requirements and record the information of this sudden decrease interval.

[0058] Combined with the 96-point curve data, identify the mutation characteristics of each effective interval, and count the number of sudden increases and sudden decreases to form the final curve feature quantity sequence.

[0059] Note: In this step of the present invention, D is a constant coefficient, the default value is 365, and it can be adjusted according to the research data source situation, and it should not be less than 30.

[0060] Step 2: The construction of the minimum interval dynamic feature curve of different electricity consumption behaviors of users described in step S2 is specifically implemented according to the following steps, as Figure 2 shown:

[0061] 2.1: Based on the daily curve electricity consumption characteristics of users in step 1, count different feature quantity sequences, and take the feature quantity sequence as the standard to count and record the curve situation under this feature quantity sequence to complete the classification and merging of curve features, and the merging rule Q x = Q y , to form the user feature classification set {Q1, Q2, Q3,..., Q x}.

[0062] 2.2: Take all the curve data under a certain feature classification Q x , obtain the electricity consumption of each effective interval of each curve, calculate the average value of the interval electricity consumption corresponding to the same time point, and generate the interval electricity consumption {q1, q2, q3,..., q 95}.

[0063] Suppose: There are 30 curves under the feature classification Qx, which are curves X1-x 30 , and the q1 (from 00:00 to 00:15) of each curve is a valid value, then:

[0064] 2.3: According to each feature classification, form the corresponding feature classification curve set:

[0065]

[0066] As shown Figure 6 in the figure, the interval power consumption curves of the same user under different feature classifications (weekdays and non-weekdays) calculated based on actual data.

[0067] 2.4: Given the initial data point (the value corresponding to the 00:00 moment) R1 (a constant greater than 0), combined with the power consumption interval sequence {q1, q2, q3,..., q 95}, generate the corresponding feature curve data {R1, R2, R3,..., R 96}.

[0068] Calculation formula: R n+1 = R n + q n , where R1 is a known quantity.

[0069] Step Three: Perform similarity matching between the daily power consumption curve described in Step S3 and the minimum interval dynamic feature curve. As Figure 3 shown in the figure, it is specifically implemented according to the following steps:

[0070] 3.1 Obtain the current power consumption indication transaction curve data of the user;

[0071] 3.2 According to the missing data points of the curve data, divide one curve data into multiple continuous curve segments, obtain the interval power consumption sequence corresponding to each curve segment, and form multiple power consumption curve segments;

[0072] 3.3 According to the effective power consumption interval of the power consumption curve segment, use the Pearson coefficient method to calculate the correlation within the corresponding time series of this curve segment and each dynamic feature curve of this user.

[0073] Explanation of the curve segment correlation calculation is as follows:

[0074] Assume the curve segment q = [q1, q,..., q 14 , and obtain the data of the same time period of a certain feature curve of this user y = [y1, y2,..., y 14

[0075] Then the calculation formula is as follows:

[0076]

[0077] Among them, ρ q，y is the required correlation coefficient, cov(q, y) is the covariance of the q and y vectors, and δ q , δ y are the standard deviations of the q and y vectors respectively. The covariance calculation formula is as follows:

[0078] ​cov(q, y) = E((q - μ q )(y - μ y ))

[0079] where E represents the expectation, and μ q and μ y represent the means of q and y respectively.

[0080] Similarly, using the above calculation method, the similarity between this curve segment and other dynamic curves can be obtained. Similarly, the similarity between other curve segments of this curve of this user and all dynamic feature curves of this user can also be obtained.

[0081]

[0082] 3.4 Synthesize the correlation between each segmented curve of this curve and each dynamic feature curve, and use the similarity weighted ratio method to obtain the similarity between the whole curve and each feature curve.

[0083] Assume that the number of valid data points of each curve segment of the current running curve is: N1, N2,..., N x , and use the weight ratio to obtain the similarity between the current running curve and the dynamic feature curve Q1.

[0084]

[0085] Similarly, using the above calculation method, the similarity between the current running curve and all other dynamic feature curves Q x can be obtained.

[0086]

[0087] 3.5 According to the result obtained in step 3.4, obtain the dynamic feature curve with the highest similarity (closest electricity consumption behavior) to this curve.

[0088] Select the maximum value in the correlation calculation result: If the similarity is greater than 0.85, it is considered a successful match, and the corresponding dynamic feature curve is the data fitting reference curve for the current running curve. If the similarity is not greater than 0.85, it is considered a failed match, and no dynamic feature curve for data fitting reference is found. For this situation, the present invention calculates according to all dynamic feature curves and gives multiple fitting values for selection and application.

[0089] Step Four: The data fitting based on the electricity consumption weight with the minimum electricity consumption interval described in step S4 is implemented as follows: Figure 4 as shown, and is specifically implemented according to the following steps:

[0090] 4.1 Based on step three, obtain the current running curve data and the corresponding optimal matching dynamic feature curve data Q x ;

[0091] 4.2 Based on the missing situation of the current operation curve data, obtain the curve segment that needs to be data-fitted in the current operation curve data (the two data points before and after the data points to be fitted. For example, in the current operation curve data, for the data point R 12 -R 32 , it is necessary to obtain R 11 、R 33 two data points). At the same time, obtain the corresponding value Q of the dynamic characteristic curve for this curve segment x (R 11 -R 33 ).

[0092] 4.3 Calculate the power consumption of each fitting data point corresponding time interval according to the electricity consumption trend distribution of the corresponding period of the reference dynamic characteristic curve.

[0093] For example:

[0094] Among them: the numerical values of each data point of the characteristic curve are known, the current operation curve data points R 11 、R 33 are known, and the only unknown is the current operation curve data point R 12 . Substitute the numerical values of the known data points into the above calculation formula, and the numerical value of the unknown data point R 12 that needs to be fitted in the current operation curve can be obtained.

[0095] Similarly, the numerical values of the data points that need to be fitted can be obtained in turn R 13 -R 32 . Similarly, when a curve contains multiple segments of data that need to be fitted, the above method can be used to obtain the numerical values corresponding to the data points of each curve segment, so as to complete the fitting of the missing data of the whole day curve. For the simulation verification results of the examples of the present invention, see the appendix Figure 7 .

[0096] As can be seen from the above, in order to solve the problem of missing data in the user's electricity consumption indication curve, a method for fitting electricity load curve data based on the minimum interval dynamic distribution is proposed. This method first constructs a mathematical model of the electricity curve characteristics based on historical data, then classifies the user's electricity consumption behavior reliably, and then uses the same type of electricity consumption characteristics to effectively fit and predict the curve of the missing data, and complete the repair of the curve data. This method has high universality and accuracy in solving the problem of missing data in the individual user's electricity curve.

[0097] The present invention provides a method for fitting power consumption load curve data based on minimum interval dynamic distribution, which is used to support grid enterprises in carrying out business such as power market transaction load forecasting, trade settlement, electricity price formulation, and agency power purchase, power user comprehensive energy efficiency management, government major analysis and decision-making, etc. This algorithm can effectively ensure the data integrity of the electric energy indication curve and the accuracy of the fitted data, and overall improve the quality of the electric energy indication curve data.

[0098] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the claims and their equivalents.

Claims

1. A data fitting method for electricity load curve based on minimum interval dynamic distribution, characterized in that, It includes the following steps: S1: Identification and classification of user electricity consumption curve characteristics; S2: Construction of the minimum interval dynamic characteristic curve of different user electricity consumption behaviors; S3: Similarity matching between the daily electricity consumption curve and the minimum interval dynamic characteristic curve; S4: Data fitting based on the electricity consumption weight with the minimum electricity consumption interval; The identification and classification of user electricity consumption curve characteristics described in step S1 are specifically implemented according to the following steps: Retrieve the electric energy indication curve of the electricity-consuming customer in the recent D days, identify the characteristics of each daily curve, and record the characteristic information of each daily curve; the characteristics are described and recorded by identifying the time sequence of data points of sudden increases and decreases in interval electricity consumption and the number of sudden increases and decreases; characteristic quantity recording sequence: Q = {M, N, {T1, T2, T3,..., T M+N}}, where M is the number of sudden increases, N is the number of sudden decreases, and T x is the xth mutation information element, including the mutation time point and the mutation description; Construct the curve characteristic information set of the recent D days: The construction of the minimum interval dynamic characteristic curve of different user electricity consumption behaviors described in step S2 is specifically implemented according to the following steps: Merge the same type of electricity consumption characteristics according to the curve characteristic parameters to form the electricity consumption characteristic classification of this user: {Q1, Q2, Q3,..., Q x}; Take all the curve data under a certain characteristic classification Q x to obtain the electricity consumption of each curve for each effective interval, calculate the weighted average of the interval electricity consumption corresponding to the same time point, and generate the interval electricity consumption {q1, q2, q3,..., q 95}; The similarity matching between the daily electricity consumption curve and the minimum interval dynamic characteristic curve described in step S3 is specifically implemented according to the following steps: Obtain the current electricity consumption indication transaction curve data of the user; According to the missing data points of the curve data, divide a curve data into multiple continuous curve segments, obtain the interval electricity consumption sequence corresponding to each curve segment, and form multiple electricity consumption curve segments; According to the effective electricity consumption interval of the electricity consumption curve segment, use the Pearson coefficient method to calculate the correlation within the corresponding time series of this curve segment and each dynamic characteristic curve of the user; Based on the correlation between each segmented curve of this curve and each dynamic characteristic curve, use the similarity weighted proportion method to calculate the similarity between the whole curve and each characteristic curve; The calculation description of the curve segment correlation is as follows: Define the curve segment q = [q1, q, … q 14 , and obtain the data y = [y1, y2, … y 14 ​ Then the calculation formula is as shown below: where ρ q,y is the required correlation coefficient, cov(q, y) is the covariance of vectors q and y, and δ q , δ y are the standard deviations of vectors q and y respectively. The covariance calculation formula is as follows: cov(q,y) = E((q - μ q )(y - μ y )) where E represents the expectation, μ q , μ y represent the means of q and y respectively.

2. The method for fitting power consumption load curve data based on minimum interval dynamic distribution according to claim 1, wherein The data fitting based on the electricity consumption weight with the minimum electricity consumption interval described in step S4 is specifically implemented according to the following steps: Obtain two credible values before and after the missing data curve segment. According to the trends of the corresponding characteristic curves in the same period and the total electricity consumption, use the weighted electricity consumption ratio to calculate the fitted electricity consumption of each interval Then, obtain the curve fitting indication value based on the credible points; where, Q: electricity consumption of the interval to be obtained; Q 总 : total electricity consumption of the missing data points of the curve; The electricity consumption corresponding to the interval electricity consumption reference curve for the interval to be obtained; The total electricity consumption corresponding to the missing data points of the curve to be obtained for the interval electricity consumption reference curve.

Images