CN116304523A - An Improved Photovoltaic Decomposition Algorithm Based on Capacity Estimation - Google Patents

Abstract

The invention discloses an improved photovoltaic decomposition algorithm based on capacity estimation, which comprises the following steps: generating a candidate capacity sequence; obtaining a capacity curve based on the candidate capacity sequence; performing capacity estimation based on the capacity curve; constructing a probability density function; generating a typical sample; performing maximum likelihood estimation to obtain a preliminary result of photovoltaic decomposition; obtaining a confidence coefficient; and correcting the preliminary result by using the confidence coefficient to obtain a final photovoltaic decomposition result. The invention corrects the original result based on the capacity estimation result, and enhances the robustness of the original photovoltaic decomposition algorithm. The method has higher capacity estimation accuracy, keeps better adaptability under the condition of containing a small amount of missing values, improves the photovoltaic decomposition algorithm, increases the calculation speed of the photovoltaic decomposition algorithm, and increases the accuracy of the photovoltaic decomposition algorithm.

Description

Improved photovoltaic decomposition algorithm based on capacity estimation

Technical Field

The invention belongs to the field of photovoltaic decomposition based on post-table payload data, and particularly relates to an improved photovoltaic decomposition algorithm based on capacity estimation.

Background

More and more distributed photovoltaics are being plugged into electrical systems, where a large number of distributed photovoltaics are installed on building roofs. These rooftop distributed photovoltaic systems are small in capacity, unstable, and the metering data is presented in the form of a "post-meter" payload, i.e., a separate meter cannot be deployed to measure the photovoltaic output power, and knowledge of the photovoltaic output power curve is critical to the operation and planning of the power distribution system, so a decoupling technique is required to break down the photovoltaic output power curve from the payload curve, which we call photovoltaic break down.

However, the existing photovoltaic decomposition methods cannot well solve the problem of photovoltaic decomposition due to respective limitations, and the research limitations are mainly as follows:

the problem of relying on the physical model of the photovoltaic array and the geographical location information. The photovoltaic output power curve is estimated by using a physical model of the photovoltaic array, and the photovoltaic output power curve is decomposed from the net load power by uniformly constructing the physical model of the photovoltaic array by analyzing the influence of external meteorological information such as irradiance, temperature, wind direction, air humidity and the like on the photovoltaic output power. However, it is extremely difficult to accurately obtain detailed geographic information of the rooftop distributed photovoltaic under each power distribution system, because this involves user privacy, and there are also problems that part of the distributed photovoltaic is not installed under the authority of the grid company, which hinders the obtaining of geographic information; on the other hand, the physical model is difficult to adaptively adjust individual differences in consideration of differences of different user photovoltaic panel models, conversion efficiency and the like.

Some work utilizes high resolution color satellite images to automatically identify the location and size of the miniature photovoltaic array. This work is effective for identifying photovoltaic arrays over a large area, however, it can only estimate the physical size of the photovoltaic array, but cannot directly estimate the capacity, as the capacity of the same size photovoltaic array may vary from one type of photovoltaic panel to another and its corresponding operating conditions.

Some data driven methods make use of differences between the payload curves of similar days of weather for photovoltaic system parameter estimation. This solves the dependence on physical models and privacy information to some extent, but this approach does not take into account the differences between consumer loads on similar days, while it requires longer time of payload data as a clustering basis for similar weather.

Therefore, in order to solve the above-mentioned problems, there is a need for a capacity estimation method and a photovoltaic decomposition strategy that rely on short-time payload data to achieve driving.

Disclosure of Invention

The invention provides an improved photovoltaic decomposition algorithm based on capacity estimation, which aims to solve the problems existing in the prior art.

The technical scheme of the invention is as follows: an improved photovoltaic decomposition algorithm based on capacity estimation, comprising the steps of:

A. generating a candidate capacity sequence;

B. obtaining a capacity curve based on the candidate capacity sequence;

C. performing capacity estimation based on the capacity curve;

D. constructing a probability density function;

E. generating a typical sample;

F. performing maximum likelihood estimation to obtain a preliminary result of photovoltaic decomposition;

G. obtaining a confidence coefficient;

H. and correcting the preliminary result by using the confidence coefficient to obtain a final photovoltaic decomposition result.

Further, the step A generates a candidate capacity sequence, and the specific process is as follows:

first, the payload power data is taken as input;

then, counting and recording a net load night power extremum and a net load daytime power extremum under a month scale;

finally, the candidate capacity sequence is formed by the addition and combination of the extreme value of the payload night power and the extreme value of the payload daytime power.

Further, step B obtains a capacity curve based on the candidate capacity sequence, which specifically includes the following steps:

firstly, sequencing the obtained candidate capacity sequences;

then, a capacity curve is established with sequence numbers and capacities.

Further, the step C performs capacity estimation based on the capacity curve, and the specific process is as follows:

firstly, obtaining a change trend of a capacity curve based on the capacity curve;

then, capacity estimation is performed using the trend of the capacity curve.

Further, the probability density function is constructed in the step D, and the specific process is as follows:

firstly, obtaining data information of part of considerable users;

then, the monthly night electricity consumption and the daytime electricity consumption of the consumption load are obtained based on the data information extraction;

finally, a probability density function is constructed based on the monthly night electricity usage and the daytime electricity usage.

Further, step E generates a typical sample, which is specifically as follows:

a representative sample is formed using a photovoltaic output power curve cluster.

Further, the maximum likelihood estimation is performed in the step F to obtain a preliminary result of the photovoltaic decomposition, and the specific process is as follows:

firstly, carrying out linear combination solution on a typical sample;

and then, obtaining a preliminary result of photovoltaic decomposition after obtaining the optimal weight.

Further, the confidence coefficient is obtained in the step G, which comprises the following specific steps:

firstly, obtaining a true value of capacity based on data information of part of considerable users;

then, obtaining a capacity estimation value of a considerable user;

and finally, obtaining a confidence coefficient by using the capacity estimation value and the true value of the capacity.

Further, the step H corrects the preliminary result by using the confidence coefficient to obtain a final photovoltaic decomposition result, which comprises the following specific steps:

firstly, obtaining a capacity value in a preliminary result;

then, a capacity estimation value obtained by a capacity estimation method is used;

then, processing the capacity estimation value and the capacity value to obtain a threshold value;

then, judging whether the capacity estimation is needed or not according to the threshold value;

and finally, correcting the part to be corrected by using the confidence coefficient.

The beneficial effects of the invention are as follows:

the invention utilizes the intermittent characteristic of photovoltaic power generation and the characteristic of stable power during unmanned use represented by consumer load to realize the rapid and accurate estimation of photovoltaic capacity under the net load.

According to the invention, a typical sample is formed by clustering the photovoltaic output power curves of considerable users, a probability density function is constructed by utilizing the correlation between night and daytime shown by the consumption load, and the optimal weight is obtained by solving the optimization problem, so that the preliminary estimation of the photovoltaic output power curves is realized.

The invention corrects the original result based on the capacity estimation result, and enhances the robustness of the original photovoltaic decomposition algorithm. The method has higher capacity estimation accuracy, keeps better adaptability under the condition of containing a small amount of missing values, improves the photovoltaic decomposition algorithm, increases the calculation speed of the photovoltaic decomposition algorithm, and increases the accuracy of the photovoltaic decomposition algorithm.

Drawings

FIG. 1 is a schematic diagram of the architecture of the present invention;

fig. 2 is a graph of capacity estimation of the present invention.

Detailed Description

The present invention will be described in detail below with reference to the drawings and examples:

as shown in fig. 1-2, an improved photovoltaic decomposition algorithm based on capacity estimation, comprising the steps of:

A. generating a candidate capacity sequence;

B. obtaining a capacity curve based on the candidate capacity sequence;

C. performing capacity estimation based on the capacity curve;

D. constructing a probability density function;

E. generating a typical sample;

F. performing maximum likelihood estimation to obtain a preliminary result of photovoltaic decomposition;

G. obtaining a confidence coefficient;

H. and correcting the preliminary result by using the confidence coefficient to obtain a final photovoltaic decomposition result.

Step A, generating a candidate capacity sequence, wherein the specific process is as follows:

first, the payload power data is taken as input;

then, counting and recording a net load night power extremum and a net load daytime power extremum under a month scale;

finally, the candidate capacity sequence is formed by the addition and combination of the extreme value of the payload night power and the extreme value of the payload daytime power.

And B, obtaining a capacity curve based on the candidate capacity sequence, wherein the specific process is as follows:

firstly, sequencing the obtained candidate capacity sequences;

then, a capacity curve is established with sequence numbers and capacities.

And C, estimating the capacity based on the capacity curve, wherein the specific process is as follows:

firstly, obtaining a change trend of a capacity curve based on the capacity curve;

then, capacity estimation is performed using the trend of the capacity curve.

And D, constructing a probability density function, wherein the specific process is as follows:

firstly, obtaining data information of part of considerable users;

then, the monthly night electricity consumption and the daytime electricity consumption of the consumption load are obtained based on the data information extraction;

finally, a probability density function is constructed based on the monthly night electricity usage and the daytime electricity usage.

Step E, generating a typical sample, wherein the specific process is as follows:

a representative sample is formed using a photovoltaic output power curve cluster.

And F, carrying out maximum likelihood estimation to obtain a preliminary result of photovoltaic decomposition, wherein the specific process is as follows:

firstly, carrying out linear combination solution on a typical sample;

and then, obtaining a preliminary result of photovoltaic decomposition after obtaining the optimal weight.

The confidence coefficient is obtained in the step G, and the specific process is as follows:

firstly, obtaining a true value of capacity based on data information of part of considerable users;

then, obtaining a capacity estimation value of a considerable user;

and finally, obtaining a confidence coefficient by using the capacity estimation value and the true value of the capacity.

And step H, correcting the preliminary result by using the confidence coefficient to obtain a final photovoltaic decomposition result, wherein the specific process is as follows:

firstly, obtaining a capacity value in a preliminary result;

then, a capacity estimation value obtained by a capacity estimation method is used;

then, processing the capacity estimation value and the capacity value to obtain a threshold value;

then, judging whether the capacity estimation is needed or not according to the threshold value;

and finally, correcting the part to be corrected by using the confidence coefficient.

Specifically, step A generates a candidate capacity sequence; step B, obtaining a capacity curve based on the candidate capacity sequence; and C, performing capacity estimation based on the capacity curve, and integrating a capacity estimation module.

Specifically, step D constructs a probability density function; step E, generating a typical sample; and F, carrying out maximum likelihood estimation to obtain a preliminary result of photovoltaic decomposition, and integrating a photovoltaic decomposition module.

Specifically, step G obtains a confidence coefficient; and step H, correcting the preliminary result by using the confidence coefficient to obtain a final photovoltaic decomposition result, and integrating capacity correction.

Specifically, the capacity estimation module performs capacity estimation based on a capacity estimation curve variation trend.

Specifically, the photovoltaic decomposition module is mainly used for carrying out photovoltaic decomposition on payload data of an unknown user by utilizing source load characteristics of an appreciable user.

Yet another embodiment

A. Generating candidate capacity sequences

For the net load data metered by the electric meter of the Internet of things, the value of the net load data is equal to the consumption load of a user minus the photovoltaic output power, as shown in a formula (1).

Considering that the user consumption load is zero, the absolute value of the net load is photovoltaic output power, and the peak value of the absolute value of the net load is equal to the peak value of the photovoltaic output power, namely photovoltaic capacity.

The real world situation has a consumer load such that the absolute value of the minimum of the net load is smaller than the value of the actual photovoltaic capacity.

Theoretically, in case the payload P takes a minimum value during the day, the consumption load should be as small as possible and the photovoltaic output power should be as large as possible. Thus, if an accurate estimate of this smaller consumer load value is made, the magnitude of light Fu Rong can be calculated by equation (1).

P＝L-G (1)

Where P represents the net load, L represents the consumer load, and G represents the photovoltaic output power.

Specifically, the present invention provides one of the following specific schemes to obtain candidate capacity sequences:

first, the payload data of each day in a month is divided by day and night,

then, absolute values of the night power minimum value and the day power minimum value are obtained as shown in expression (2).

Wherein D represents the number of days, T _d 、T _n Represents the time point of day on day d and the time point of night on day n, respectively, P _h And (t) represents the value of the net load power at hour t.

P _day,p (d) I.e. all minima of the payload daytime, represent photovoltaic power generation in the presence of minimal consumer load disturbance or weather conditions being optimal.

P _night,p And (n) is an estimation of a smaller consumption load value, and represents the usual power of the electric appliance in the night unmanned use state.

The additive combination of the two forms a candidate sequence set with capacity, as shown in formula (3).

C _candi (i)＝{P _day,p (d)+P _night,p (n)|d∈D,n∈D},i＝1,...,D ² (3)

B. Obtaining a capacity curve based on candidate capacity sequences

And counting the occurrence frequency of data by using a proper method by utilizing the generated candidate capacity sequence, so that rough estimation of the capacity can be realized.

However, this estimation method is theoretically smaller than the real capacity value, because the capacity value of the photovoltaic represents the peak value of the photovoltaic output power curve and is slightly larger than the point with highest occurrence frequency in the extremum of the power curve under the condition of considering only the photovoltaic effect.

The present invention provides a concept to implement capacity estimation to solve this problem:

sequencing the candidate capacity sequences, drawing a corresponding capacity estimation curve, and estimating the capacity value by utilizing the change trend of the curve as shown by a solid line in fig. 2, wherein the steps are as follows:

a. the maximum point (i.e., the last point) of the capacity estimation curve is marked as the termination point.

b. Find the first point where the value is greater than the absolute value of the minimum of the payload (as shown in equation (4)) as the starting point,

c. and taking the capacity value sequence between the starting point and the ending point as a capacity sequence to be determined.

d. A straight line connecting the starting point and the ending point is made, a tangent line parallel to the straight line is found out about the curve of the capacity sequence segment to be determined, and the tangent point is the estimated value of the capacity

C. Performing capacity estimation based on the capacity curve;

the first three steps described above determine the upper and lower limits of the estimated capacity. In the absence of measurement errors, the capacity value must be greater than the absolute value P of the minimum value of the payload due to the presence of the consumer load _floor This is used as a lower limit of the capacity estimation value.

The overall slope of the curve between the starting point and the tangent point is obviously smaller than that between the tangent point and the ending point, because the 'stable section' between the starting point and the tangent point represents the section with the highest occurrence frequency of the capacity value, which is also the section with the highest probability density of the whole capacity sequence section to be determined, and the self-adaptive increment is obtained by selecting the tangent point to ensure that the value with the highest occurrence probability of the section to realize capacity estimation.

D. Construction of probability Density function

At the monthly scale, there is a high correlation between the total amount of consumer load in the daytime and the total amount of consumer load in the nighttime. The night time situation is when no photovoltaic power generation is generated, the net load data is equal to the consumption load data, and the daytime consumption load can be estimated by using the correlation between the night time consumption load and the daytime consumption load, namely, the correlation between the night time net load and the daytime consumption load.

This step uses this correlation to construct a joint probability density function of daytime and nighttime monthly consumption loads.

First, the daily consumption load and the night consumption load are respectively accumulated to obtain the monthly consumption load, as shown in a formula (5).

Then, a gaussian mixture model (Gaussian mixture modeling, GMM) is used to construct a monthly night and daytime consumption load joint distribution of the considerable user, as shown in the following formula (6).

Where f (·, ·) represents the joint probability density function estimated using GMM, Λ= { S, θ _k ,μ _k ,Σ _k All are parameters in GMM that need to be learned with the data of the considerable user.

The set of solution parameters is converted into an optimization problem, as shown in equation (7), and the solution is performed using an expectation-maximization (EM) algorithm.

Wherein N isRepresents the total number of users, L _day,m (j)、L _night,m (j) Representing the sum of the month daytime power and the sum of the night power of the jth user respectively.

E. Generating a representative sample

Considering that the weather conditions accepted by different users are highly similar within a defined geographical range, the photovoltaic output power curves they produce due to the photovoltaic power generation mechanism are also highly similar, so that the photovoltaic output power curves of unknown users can be represented by linear combinations of the photovoltaic output power curves of the considerable users.

The invention clusters the photovoltaic output power curves known to a considerable user to form a typical sample so as to improve the calculation efficiency.

All known photovoltaic output power curves are first normalized as shown in equation (8).

Wherein G is _h (t) represents a photovoltaic output power curve, G _max And G _min Representing the maximum and minimum of the photovoltaic output power curve, respectively.

Then, carrying out mean shift clustering on all normalized standard photovoltaic output power curves, and selecting the cluster centers of the generated N clusters as typical samples

And calculate its current month power sum

F. Performing maximum likelihood estimation to obtain a preliminary result of photovoltaic decomposition

With the net load equal to the consumer load subtracted from the photovoltaic power generation, we can derive equation (9) for the monthly cumulative sum.

Wherein w= [ w ] ₁ ,…w _N ] ^T Representing the unknown weights that need to be solved.

In combination with probability density functions that have been generated by a gaussian mixture model, a constrained optimization problem of the following formula (10) is constructed:

the physical meaning of the constraint condition is that the photovoltaic power generation power is positive, and the consumption load is positive.

The optimization problem can be solved through a numerical calculation method to obtain optimal weights, and then the weights are weighted

Endowing a photovoltaic output power curve with a typical sample +.>

Performing linear combination to obtain preliminary estimation result +.>

G. Obtaining confidence coefficient

In practice, not all users can strictly meet the data distribution generated by the GMM. The whole deviation of the photovoltaic curve estimation result is caused by the fact that partial users do not meet the data distribution, the deviation is larger than the real result, and the capacity correction post-processing is carried out on the preliminary result of the photovoltaic decomposition in consideration of the specificity of the users.

Firstly, performing unsupervised capacity estimation on a considerable user to obtain an estimated value

And comparing the true capacity value C, and calculating to obtain the ratio as a confidence coefficient, wherein the confidence coefficient is shown in a formula (11).

Where N represents the number of considerable users.

H. Correcting the preliminary result by using the confidence coefficient to obtain a final photovoltaic decomposition result

Capacity estimation of unknown user

Then carrying out photovoltaic decomposition to obtain a preliminary result +.>

Peak of (2)

Judging whether to correct by comparing the ratio of the two with the confidence coefficient, and correcting the photovoltaic decomposition result meeting the judgment condition of the formula (12) by the formula (13) to obtain a final photovoltaic decomposition result +.>

The invention utilizes the intermittent characteristic of photovoltaic power generation and the characteristic of stable power during unmanned use represented by consumer load to realize the rapid and accurate estimation of photovoltaic capacity under the net load.

According to the invention, a typical sample is formed by clustering the photovoltaic output power curves of considerable users, a probability density function is constructed by utilizing the correlation between night and daytime shown by the consumption load, and the optimal weight is obtained by solving the optimization problem, so that the preliminary estimation of the photovoltaic output power curves is realized.

The invention corrects the original result based on the capacity estimation result, and enhances the robustness of the original photovoltaic decomposition algorithm. The method has higher capacity estimation accuracy, keeps better adaptability under the condition of containing a small amount of missing values, improves the photovoltaic decomposition algorithm, increases the calculation speed of the photovoltaic decomposition algorithm, and increases the accuracy of the photovoltaic decomposition algorithm.

Claims (9)

1. An improved photovoltaic decomposition algorithm based on capacity estimation, characterized in that: the method comprises the following steps:

(A) Generating a candidate capacity sequence;

(B) Obtaining a capacity curve based on the candidate capacity sequence;

(C) Performing capacity estimation based on the capacity curve;

(D) Constructing a probability density function;

(E) Generating a typical sample;

(F) Performing maximum likelihood estimation to obtain a preliminary result of photovoltaic decomposition;

(G) Obtaining a confidence coefficient;

(H) And correcting the preliminary result by using the confidence coefficient to obtain a final photovoltaic decomposition result.

2. An improved photovoltaic decomposition algorithm based on capacity estimation according to claim 1, wherein: step (A) generating a candidate capacity sequence, wherein the specific process is as follows:

first, the payload power data is taken as input;

then, counting and recording a net load night power extremum and a net load daytime power extremum under a month scale;

finally, the candidate capacity sequence is formed by the addition and combination of the extreme value of the payload night power and the extreme value of the payload daytime power.

3. An improved photovoltaic decomposition algorithm based on capacity estimation according to claim 1, wherein: step (B) obtaining a capacity curve based on the candidate capacity sequence, wherein the specific process is as follows:

firstly, sequencing the obtained candidate capacity sequences;

then, a capacity curve is established with sequence numbers and capacities.

4. An improved photovoltaic decomposition algorithm based on capacity estimation according to claim 1, wherein: and (C) estimating the capacity based on the capacity curve, wherein the specific process is as follows:

firstly, obtaining a change trend of a capacity curve based on the capacity curve;

then, capacity estimation is performed using the trend of the capacity curve.

5. An improved photovoltaic decomposition algorithm based on capacity estimation according to claim 1, wherein: and (D) constructing a probability density function, wherein the specific process is as follows:

firstly, obtaining data information of part of considerable users;

then, the monthly night electricity consumption and the daytime electricity consumption of the consumption load are obtained based on the data information extraction;

finally, a probability density function is constructed based on the monthly night electricity usage and the daytime electricity usage.

6. An improved photovoltaic decomposition algorithm based on capacity estimation according to claim 1, wherein: step (E) generates a typical sample, which comprises the following specific steps:

a representative sample is formed using a photovoltaic output power curve cluster.

7. An improved photovoltaic decomposition algorithm based on capacity estimation according to claim 1, wherein: and (F) carrying out maximum likelihood estimation to obtain a preliminary result of photovoltaic decomposition, wherein the specific process is as follows:

firstly, carrying out linear combination solution on a typical sample;

and then, obtaining a preliminary result of photovoltaic decomposition after obtaining the optimal weight.

8. An improved photovoltaic decomposition algorithm based on capacity estimation according to claim 1, wherein: the confidence coefficient is obtained in the step (G), and the specific process is as follows:

firstly, obtaining a true value of capacity based on data information of part of considerable users;

then, obtaining a capacity estimation value of a considerable user;

and finally, obtaining a confidence coefficient by using the capacity estimation value and the true value of the capacity.

9. An improved photovoltaic decomposition algorithm based on capacity estimation according to claim 1, wherein: correcting the preliminary result by using the confidence coefficient to obtain a final photovoltaic decomposition result, wherein the specific process is as follows:

firstly, obtaining a capacity value in a preliminary result;

then, a capacity estimation value obtained by a capacity estimation method is used;

then, processing the capacity estimation value and the capacity value to obtain a threshold value;

then, judging whether the capacity estimation is needed or not according to the threshold value;

and finally, correcting the part to be corrected by using the confidence coefficient.

Images