A short-term power forecasting method for photovoltaic power generation based on linear estimation method

Abstract

The invention discloses a photovoltaic power generation short-term power prediction method based on a linear estimation method, comprising: establishing a photovoltaic panel output linear model; extending the photovoltaic panel output linear model to t samples from time 1 to time t; based on the least squares criterion, obtaining a parameter estimation calculation formula for performing parameter estimation on the photovoltaic short-term power output linear model based on full data; establishing a recursive prediction model for the parameter estimation value of the linear model using only the instantaneous observation data of the power influence factor and the instantaneous observation value of the active power at the previous moment; recursively estimating the parameter estimation value of the linear model using only the instantaneous observation data of the power influence factor and the instantaneous observation value of the active power at the previous moment, effectively solving the problems of large amount of calculation and low calculation efficiency caused by parameter estimation based on full data, and having the advantages of real-time parameter estimation and online estimation.

Description

Technical Field

[0001] The invention relates to a method for predicting short-term power of power generation, and in particular to a method for predicting short-term power of photovoltaic power generation based on a linear prediction method. Background Art

[0002] When using full data to estimate the parameters of the linear model of photovoltaic power generation short-term power output, the linear estimation method and full historical data are used for estimation. As time goes by, the sample size of historical data from time t1 to time tn continues to increase, resulting in a larger amount of calculation and lower model training efficiency. Summary of the invention

[0003] In order to address the deficiencies of the prior art, the present invention aims to solve the problems of large amount of calculation, low calculation efficiency and inability to meet the real-time training requirements of the model when using full data to estimate the parameters of the photovoltaic short-term power output linear model.

[0004] In order to achieve the above-mentioned object, the present invention relates to: 1. A photovoltaic power generation short-term power prediction method based on a linear prediction method, characterized in that it comprises the following steps:

[0005] Step 1: Establish a linear model of photovoltaic panel output;

[0006] Step 2: Expand the photovoltaic panel output linear model in step 1 to t samples from time 1 to time t;

[0007] Step 3: Based on the least squares criterion, a parameter estimation formula is obtained for estimating the parameters of the photovoltaic short-term power output linear model based on the full amount of data;

[0008] Step 4: Using only the instantaneous observation data of the power influencing factor and the instantaneous observation value of the active power at the previous moment, a recursive prediction model for the parameter estimation value of the linear model is established.

[0009] Furthermore, the method for establishing the photovoltaic panel output linear model in step 1 is:

[0010] The photovoltaic panel output model refers to the functional relationship between the instantaneous value of the active power of the photovoltaic panel and the instantaneous value of the power influencing factor at the same time. Under natural conditions without human interference, the instantaneous value of the active power of the photovoltaic panel at a certain moment y and the instantaneous values ​​of the m power influencing factors at that moment x1, x2, ..., x m It is a linear relationship, that is, the output model is the following linear model

[0011]

[0012] where a0, a1, a2, …, a mare the parameters of m power impact factors, X is the power impact factor matrix, and θ is the parameter matrix.

[0013] Furthermore, the m power influencing factors include irradiance, temperature, humidity and wind speed.

[0014] Furthermore, the specific method of step 2 is:

[0015] Expanding formula (1) to t samples from time 1 to time t, we have the following matrix expression:

[0016] Y t =A t ·θ t (2)

[0017] In the above matrix expression, the subscript of each symbol is t, which means that the value of the variable is related to the historical data from time 1 to time t, where

[0018]

[0019] Furthermore, the step 3 comprises the following steps:

[0020] Step 3.1: According to the least squares criterion, the parameter θ t Estimated value of The residual sum of squares should be minimized. The residual sum of squares is the sum of squares of the differences between the short-term power observations and the predicted values ​​of all samples in the historical data. The calculation method of the residual sum of squares is as follows:

[0021]

[0022] Step 3.2: θ t Estimated value of Q(θ t ) obtains the minimum value, since the function Q(θ t ) is differentiable at any order, so θ t Estimated value of is Q(θ t ), that is, the point where the partial derivative is 0, that is, is the following system of equations:

[0023]

[0024] Substitute equation (3) in step 3.1 into equation (4) and further simplify to obtain:

[0025] A t T A t θ t =A t T Yt

[0026] When the matrix A t T A t When the rank is m+1, the above system of equations has a unique solution

[0027]

[0028] Formula (5) is based on the total data Y t ,A t The parameter estimation formula is used to estimate the parameters of the photovoltaic short-term power output linear model.

[0029] Furthermore, the step 4 specifically includes:

[0030] Step 4.1: Parameter estimation initialization: Initialize the model based on historical data from time 1 to time m+1 (t≥m+1).

[0031] From (5), we can get

[0032]

[0033] make

[0034]

[0035] in

[0036]

[0037] Step 4.2: Give the recursive formula. For any time n (m+2≤n≤t), firstly according to formula (5):

[0038]

[0039] make but

[0040]

[0041]

[0042] so

[0043]

[0044] Considering the representativeness of the subscript n, we can get the following formula

[0045]

[0046] Multiply both sides of the above formula by Conclusion

[0047]

[0048] so

[0049]

[0050] The above formula (6) is A recursive prediction model for the parameter estimates of the linear model.

[0051] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:

[0052] (1) The present invention utilizes a linear model parameter estimation recursive method, and only needs to use the instantaneous observation data of the power influence factor and the instantaneous observation value of the active power at the previous moment to recursively estimate the parameter estimation value of the linear model, which effectively solves the problems of large amount of calculation and low calculation efficiency caused by parameter estimation based on the full amount of data, and realizes the real-time training of the linear model; thereby improving the parameter estimation efficiency, and having the advantages of real-time parameter estimation and online estimation. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in each embodiment of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0054] The present invention relates to a method for short-term photovoltaic power generation prediction based on a linear estimation method, comprising the following steps:

[0055] Step 1: Establish a linear model of photovoltaic panel output; the photovoltaic panel output model refers to the functional relationship expression between the instantaneous value of the active power of the photovoltaic panel and the instantaneous value of the power influencing factor at the same time. Under natural conditions without human interference, the instantaneous value of the active power y of the photovoltaic panel at a certain moment and the instantaneous values ​​x1, x2, ..., x m It is a linear relationship, that is, the output model is the following linear model

[0056]

[0057] where a0, a1, a2, …, a m are the parameters of m power influencing factors (including irradiance, temperature, humidity and wind speed), X is the power influencing factor matrix, and θ is the parameter matrix.

[0058] Step 2: Expand the linear model of photovoltaic panel output in step 1 to t samples from time 1 to time t; the specific method is:

[0059] Expanding formula (1) to t samples from time 1 to time t, we have the following matrix expression:

[0060] Y t =A t ·θ t (2)

[0061] In the above matrix expression, the subscript of each symbol is t, which means that the value of the variable is related to the historical data from time 1 to time t, where

[0062]

[0063] Step 3: Based on the least squares criterion, a parameter estimation formula is obtained for estimating the parameters of the photovoltaic short-term power output linear model based on the full amount of data;

[0064] The specific steps include:

[0065] Step 3.1: According to the least squares criterion, the parameter θ t Estimated value of The residual sum of squares should be minimized. The residual sum of squares is the sum of the squares of the differences between the short-term power observations and the predicted values ​​of all samples in the historical data. Intuitively, a good prediction should make the short-term power predictions and the observed values ​​as close as possible on as many samples as possible, that is, the residual sum of squares should be as small as possible. The calculation method of the residual sum of squares is as follows:

[0066]

[0067] Step 3.2: θ t Estimated value of Q(θ t ) obtains the minimum value, since the function Q(θ t ) is differentiable at any order, so θ t Estimated value of is Q(θ t ), that is, the point where the partial derivative is 0, that is, is the following system of equations:

[0068]

[0069] Substitute equation (3) in step 3.1 into equation (4) and further simplify to obtain:

[0070] A t T A t θ t =A t T Yt

[0071] When the matrix A t T A t When the rank is m+1, the above system of equations has a unique solution

[0072]

[0073] Formula (5) is based on the total data Y t ,A t The parameter estimation formula is used to estimate the parameters of the photovoltaic short-term power output linear model.

[0074] In step 4 and formula (5), the parameters are estimated based on the full data, which has serious performance problems: in the production environment, sample data is acquired incrementally. As data continues to accumulate, the full data Y t ,A t The scale of the power influencing factor is getting larger and larger, resulting in an increasing amount of calculation for parameter estimation and lower calculation efficiency, which makes it difficult to meet the real-time calculation requirements in the production environment. The following is a method for establishing a recursive prediction model for the parameter estimation value of the linear model using only the instantaneous observation data of the power influencing factor and the instantaneous observation value of the active power at the previous moment:

[0075] Step 4.1: Parameter estimation initialization: Initialize the model based on historical data from time 1 to time m+1 (t≥m+1).

[0076] From (5), we can get

[0077]

[0078] make

[0079]

[0080] in

[0081]

[0082] Step 4.2: Give the recursive formula. For any time n (m+2≤n≤t), firstly according to formula (5):

[0083]

[0084] make but

[0085]

[0086]

[0087]

[0088] so

[0089]

[0090] Considering the representativeness of the subscript n, we can get the following formula

[0091]

[0092] Multiply both sides of the above formula by Conclusion

[0093]

[0094] so

[0095]

[0096] The above formula (6) is A recursive prediction model for the parameter estimates of the linear model.

[0097] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A photovoltaic power generation short-term power prediction method based on a linear estimation method, characterized in that: The steps include: Step 1: Establish a linear model of photovoltaic panel output; Step 2: Expand the photovoltaic panel output linear model in step 1 to t samples from time 1 to time t; Step 3: Based on the least squares criterion, a parameter estimation formula is obtained for estimating the parameters of the photovoltaic short-term power output linear model based on the full amount of data; Step 4: Using only the instantaneous observation data of the power influence factor and the instantaneous observation value of the active power at the previous moment, a recursive prediction model for the parameter estimation value of the linear model is established; The method of establishing the photovoltaic panel output linear model in step 1 is: The photovoltaic panel output model refers to the functional relationship between the instantaneous value of the active power of the photovoltaic panel and the instantaneous value of the power influencing factor at the same time. Under natural conditions without human interference, the instantaneous value of the active power of the photovoltaic panel at a certain moment y and the instantaneous values ​​of the m power influencing factors at that moment x1, x2, ..., x m It is a linear relationship, that is, the output model is the following linear model where a0, a1, a2, …, a m are the parameters of m power impact factors, X is the power impact factor matrix, and θ is the parameter matrix; The specific method of step 2 is: Expanding formula (1) to t samples from time 1 to time t, we have the following matrix expression: Y t =A t ·i t (2) In the above matrix expression, the subscript of each symbol is t, which means that the value of the variable is related to the historical data from time 1 to time t, where The step 3 comprises the following steps: Step 3.1: According to the least squares criterion, the parameter θ t Estimated value of The residual sum of squares should be minimized. The residual sum of squares is the sum of squares of the differences between the short-term power observations and the predicted values ​​of all samples in the historical data. The calculation method of the residual sum of squares is as follows: Step 3.2: θ t Estimated value of Q(θ t ) obtains the minimum value, since the function Q(θ t ) is differentiable at any order, so θ t Estimated value of is Q(θ t ), that is, the point where the partial derivative is 0, that is, is the following system of equations: Substitute equation (3) in step 3.1 into equation (4) and further simplify to obtain: A t T A t i t =A t T Y t When the matrix A t T A t When the rank is m+1, the above system of equations has a unique solution (5) Formula (5) is based on the total data Y t ,A t The parameter estimation formula is used to estimate the parameters of the photovoltaic short-term power output linear model.

2. The photovoltaic power generation short-term power prediction method based on the linear estimation method according to claim 1 is characterized in that: The m power influencing factors include irradiance, temperature, humidity and wind speed.

3. The photovoltaic power generation short-term power prediction method based on the linear estimation method according to claim 1 is characterized in that: The step 4 specifically includes: Step 4.1: Parameter estimation initialization: Initialize the model based on historical data from time 1 to time m+1 (t≥m+1). From (5), we can get make in Step 4.2: Give the recursive formula. For any time n (m+2≤n≤t), firstly according to formula (5): make but so Considering the representativeness of the subscript n, we can get the following formula Multiply both sides of the above formula by Conclusion so The above formula (6) is A recursive prediction model for the parameter estimates of the linear model.

Images