CN115271198A - Net load prediction method and device of photovoltaic equipment - Google Patents

Abstract

The invention relates to the technical field of photovoltaic net load prediction of power systems, and specifically provides a method and device for predicting the net load of photovoltaic equipment, including: using a pre-built regression model to predict the wavelet data of the net load of photovoltaic equipment in a high-dimensional feature space. Sparse solution; perform wavelet reconstruction on the sparse solution of the wavelet data of the net load of the photovoltaic equipment in the high-dimensional feature space, and obtain the net load prediction data of the photovoltaic equipment; wherein, the wavelet data includes: low frequency coefficients and high frequency coefficients . The technical solution provided by the present invention can effectively realize the dimensionality reduction of data and the high-precision prediction of photovoltaic load, can realize the analysis of various external characteristics and power consumption behavior, realize the selection of high-dimensional data variables and high-precision prediction, so as to make reasonable arrangements The operation plan of the power system.

Description

Net load forecasting method and device for photovoltaic equipment

technical field

The invention relates to the technical field of photovoltaic net load forecasting of electric power systems, in particular to a method and device for net load forecasting of photovoltaic equipment.

Background technique

The net load of the distribution network refers to the difference between the user's electricity load and the renewable energy generation load, that is, the load value provided by the main grid of the power system to the distribution network. Here, it refers specifically to the household's electricity load and the utilization of household photovoltaic equipment The difference between the conversion of sunlight into electricity. At present, there are relatively few analyzes and forecasts for net load data. There are two methods for net load forecasting. One is to predict electricity load and renewable energy generation separately, and then calculate the difference, which is indirect forecasting. The load history series selects the appropriate model for forecasting, that is, direct forecasting. According to the classification of forecasting results, load forecasting can be divided into short-term load forecasting and medium- and long-term load forecasting. For short-term load forecasting, domestic and foreign research methods are roughly divided into two categories: one is the traditional method represented by the time series method; the other is the new artificial intelligence method represented by the artificial neural network.

The former includes time series method, regression analysis method and wavelet analysis and forecasting method, among which time series method is one of the classic forecasting methods of short-term load forecasting. The random variable in the model, according to the statistical regularity to study the change process of the random variable, derives the expression to predict the load, the method can be roughly divided into the following processes: autoregressive (AR) process; sliding-average (MA ) process; autoregressive moving average (ARMA) process; integral type autoregressive sliding (ARIMA) process; sequence modeled with transfer function (TF). The classic time series method has relatively less demand for historical load data, and does not require manual intervention, so the workload is relatively small. The calculation speed of the entire forecasting process is excellent and can be completed automatically. These are the advantages of this method. The disadvantage is that it relies too much on power load data and does not deal with other variable factors enough, resulting in the inability to achieve relatively high prediction accuracy. Generally, it is only suitable for relatively stable and uniform load prediction. The regression analysis method is also one of the classic forecasting methods based on the historical data of the load, but the regression analysis method also adds consideration of the external factors that affect the load on the basis of the historical data. Its advantage is that it is simple and convenient, and its disadvantage is that it cannot fully reflect the changes between the actual load and the influencing factors. The wavelet analysis and prediction method is to distinguish the loads of different properties by selecting and classifying the wavelets, and then select a certain load for different properties, analyze its laws to determine the corresponding prediction method, and predict the selected loads separately. , after the sequence is decomposed, the sequence is reconstructed, and finally the purpose of prediction is achieved. The wavelet prediction method can observe the details of the signal, especially some singular signals in the signal, and its response is particularly sensitive, and it can handle sudden changes or some weak signals very well.

The latter currently mainly uses machine learning and deep learning methods for load forecasting, and can more accurately predict nonlinear and non-stationary data. Among them, artificial neural network (ANN) and recurrent neural network (RNN) are typical representatives of this kind of method. There are many studies on the application of ANN theory to short-term load forecasting. Its outstanding advantages are that it has an adaptive function for a large number of non-structural and inaccurate laws, and has the characteristics of information memory, autonomous learning, knowledge reasoning and optimal calculation. ANN has strong self-learning and complex nonlinear function fitting capabilities, and is very suitable for power load forecasting problems. However, the research process also shows that the ANN method has local optimum, large generalization error, and difficult to determine the number of hidden units. question. Through the introduction of the concept of timing, the RNN model has a great advantage over traditional neural networks. It can store the information learned at the previous moment inside the network, and can use the information at the previous moment when processing the data at each moment. , the information transmission is continuous, so that RNN can well deal with the problems of periodic data modeling such as time series. However, when using information that is far from the current moment, RNN often cannot make full use of this information. This is the gradient disappearance problem of RNN. In practice, RNN can only process information within a very limited time. In order to solve the problem of RNN gradient disappearance, variants of RNN such as long short-term memory neural network (LSTM) and gated recurrent unit (GRU) have been proposed one after another. Therefore, variants of LSTM such as Q-RNN were subsequently proposed. The above algorithm was adopted It has been proved that it has better accuracy or faster speed than RNN in load forecasting. From the above description, it can be seen that different prediction methods have different application scopes and performance advantages. In addition, since the net load is affected by factors such as temperature and light, it is impossible to obtain more accurate results by using a single algorithm to predict the net load.

Contents of the invention

In order to overcome the above defects, the present invention proposes a net load forecasting method and device for photovoltaic equipment.

In the first aspect, a net load forecasting method of photovoltaic equipment is provided, and the net load forecasting method of photovoltaic equipment includes:

Sparse solution of wavelet data in high-dimensional feature space to predict net load of photovoltaic equipment using pre-built regression model;

Performing wavelet reconstruction on the sparse solution of the wavelet data of the net load of the photovoltaic equipment in the high-dimensional feature space, to obtain the net load prediction data of the photovoltaic equipment;

Wherein, the wavelet data includes: low-frequency coefficients and high-frequency coefficients.

Preferably, the calculation formula of the low-frequency coefficient of the net load of the photovoltaic device is as follows:

The calculation formula of the high frequency coefficient of the net load of the photovoltaic equipment is as follows:

为

对应的分解系数，

为对应缩放常数m和平移常数n选择的小波函数，t为当前时刻，D为光伏设备的净负荷的高频系数，

为ψ_mn对应的分解系数，ψ_mn为与

的互补的小波函数。In the above formula, A is the low frequency coefficient of the net load of the photovoltaic equipment,

for

The corresponding decomposition coefficient,

is the wavelet function selected corresponding to the scaling constant m and the translation constant n, t is the current moment, D is the high frequency coefficient of the net load of the photovoltaic equipment,

is the decomposition coefficient corresponding to ψ _mn , and ψ _mn is the

The complementary wavelet function of .

对应的分解系数的计算式如下：Further, the

The calculation formula of the corresponding decomposition coefficient is as follows:

The calculation formula of the decomposition coefficient corresponding to the ψ _mn is as follows:

In the above formula, T is the net load sequence length of photovoltaic equipment, p _t is the net load of photovoltaic equipment at time t.

Preferably, the acquisition process of the pre-built regression model includes:

Perform wavelet transformation on the historical net load data of the photovoltaic equipment to obtain the wavelet data of the historical net load data of the photovoltaic equipment;

The wavelet data of the historical net load data of the photovoltaic equipment is extended to the high-dimensional feature space by using the nuclear method, and the linear data corresponding to the wavelet data of the historical net load data of the photovoltaic equipment are obtained;

Using the linear data to construct training data and verification data;

Utilize the training data to train the initial Lasso regression model, utilize the verification data to verify the trained Lasso regression model, until the model index of the trained Lasso regression model meets the convergence condition, obtain the regression model constructed in advance .

Further, in the process of using the training data to train the initial regression model, the calculation formula of the regression model loss function is as follows:

L ^Reg (β) = L ^OLS (β) + P

In the above formula, L ^Reg is the penalty loss function, L ^OLS is the standard loss function, P is the penalty function value, and β is the regression coefficient vector.

Further, the calculation formula of the standard loss function is as follows:

LOLS (β)＝||Y- ^Xβ || ²

In the above formula, Y is an x×1-dimensional predictor variable matrix, X is an x×p-dimensional outcome variable vector, x is the number of observations, and p is the number of predictors.

Further, the model index includes at least one of the following: mean absolute percentage error, mean square error.

Further, the formula for calculating the mean absolute percentage error is as follows:

The formula for calculating the mean square error is as follows:

为所述验证数据中第i个样本数据的预测值，MAPE为均方误差。In the above formula, MSE is the mean absolute percentage error, I is the total number of sample data in the verification data, and y _i is the actual value of the i-th sample data in the verification data,

is the predicted value of the i-th sample data in the verification data, and MAPE is the mean square error.

In a second aspect, a net load forecasting device for photovoltaic equipment is provided, and the net load forecasting device for photovoltaic equipment includes:

The prediction module is used to predict the sparse solution of the wavelet data of the net load of the photovoltaic equipment in the high-dimensional feature space by using the regression model constructed in advance;

The reconstruction module is used to perform wavelet reconstruction on the sparse solution of the wavelet data of the net load of the photovoltaic equipment in the high-dimensional feature space, and obtain the net load prediction data of the photovoltaic equipment;

Wherein, the wavelet data includes: low-frequency coefficients and high-frequency coefficients.

Preferably, in the prediction module, the calculation formula of the low frequency coefficient of the net load of the photovoltaic equipment is as follows:

The calculation formula of the high frequency coefficient of the net load of the photovoltaic equipment is as follows:

为

对应的分解系数，

为对应缩放常数m和平移常数n选择的小波函数，t为当前时刻，D为光伏设备的净负荷的高频系数，

为ψ_mn对应的分解系数，ψ_mn为与

的互补的小波函数。In the above formula, A is the low frequency coefficient of the net load of the photovoltaic equipment,

for

The corresponding decomposition coefficient,

is the wavelet function selected corresponding to the scaling constant m and the translation constant n, t is the current moment, D is the high frequency coefficient of the net load of the photovoltaic equipment,

is the decomposition coefficient corresponding to ψ _mn , and ψ _mn is the

The complementary wavelet function of .

对应的分解系数的计算式如下：Further, the

The calculation formula of the corresponding decomposition coefficient is as follows:

The calculation formula of the decomposition coefficient corresponding to the ψ _mn is as follows:

In the above formula, T is the net load sequence length of photovoltaic equipment, p _t is the net load of photovoltaic equipment at time t.

Preferably, in the prediction module, the acquisition process of the pre-built regression model includes:

Perform wavelet transformation on the historical net load data of the photovoltaic equipment to obtain the wavelet data of the historical net load data of the photovoltaic equipment;

Using nuclear devices to expand the wavelet data of the historical net load data of photovoltaic equipment to high-dimensional feature space, and obtain the linear data corresponding to the wavelet data of the historical net load data of photovoltaic equipment;

Using the linear data to construct training data and verification data;

Utilize the training data to train the initial Lasso regression model, utilize the verification data to verify the trained Lasso regression model, until the model index of the trained Lasso regression model meets the convergence condition, obtain the regression model constructed in advance .

Further, in the process of using the training data to train the initial regression model, the calculation formula of the regression model loss function is as follows:

L ^Reg (β) = L ^OLS (β) + P

In the above formula, L ^Reg is the penalty loss function, LOLS is the standard loss function, P is the penalty function value, and β is the regression coefficient vector.

Further, the calculation formula of the standard loss function is as follows:

LOLS (β)＝||Y- ^Xβ || ²

In the above formula, Y is an x×1-dimensional predictor variable matrix, X is an x×p-dimensional outcome variable vector, x is the number of observations, and p is the number of predictors.

Further, the model index includes at least one of the following: mean absolute percentage error, mean square error.

Further, the formula for calculating the mean absolute percentage error is as follows:

The formula for calculating the mean square error is as follows:

为所述验证数据中第i个样本数据的预测值，MAPE为均方误差。In the above formula, MSE is the mean absolute percentage error, I is the total number of sample data in the verification data, and y _i is the actual value of the i-th sample data in the verification data,

is the predicted value of the i-th sample data in the verification data, and MAPE is the mean square error.

In a third aspect, a computer device is provided, including: one or more processors;

The processor is configured to store one or more programs;

When the one or more programs are executed by the one or more processors, the net load forecasting method of the photovoltaic device is realized.

In a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed, the above method for net load forecasting of a photovoltaic device is implemented.

The above-mentioned one or more technical solutions of the present invention have at least one or more of the following beneficial effects:

The present invention provides a net load prediction method and device for photovoltaic equipment, including: using a pre-built regression model to predict the sparse solution of the wavelet data of the net load of photovoltaic equipment in a high-dimensional feature space; The sparse solution of the wavelet data in the high-dimensional feature space is subjected to wavelet reconstruction to obtain the net load prediction data of the photovoltaic equipment; wherein, the wavelet data includes: low-frequency coefficients and high-frequency coefficients. The technical solution provided by the present invention can effectively realize data dimensionality reduction and high-precision prediction of photovoltaic load, and can realize analysis of various external characteristics and power consumption behavior, realize high-dimensional data variable screening and high-precision prediction, and thus rationally arrange The operation plan of the power system;

Further, the technical solution provided by the present invention introduces a prediction model combining wavelet transform with Lasso regression model, in which wavelet transform reverses the time-frequency domain of time series data and focuses on the details of the data, which is more suitable for describing the photovoltaic net load Intrinsic characteristics, and the kernel method is introduced in the Lasso regression model to map the original data to a suitable high-dimensional feature space, so that the Lasso regression model can be applied to the nonlinear photovoltaic net load data, and finally realize the accurate prediction of photovoltaic load and maintain the stability of the large power grid. Stable operation.

Description of drawings

Fig. 1 is the schematic flow chart of main steps of the net load forecasting method of the photovoltaic equipment of the embodiment of the present invention;

Fig. 2 is a main structural block diagram of a net load forecasting device for photovoltaic equipment according to an embodiment of the present invention.

Detailed ways

The specific implementation manners of the present invention will be further described in detail below in conjunction with the accompanying drawings.

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments It is a part of embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

As disclosed in the background technology, in order to achieve the double carbon target and alleviate the current shortage of electricity in various places, the use of household photovoltaic equipment has been promoted in many regions, and families with photovoltaic equipment will use photovoltaic equipment under sunlight. When the conditions are right, the self-sufficiency of electricity can be achieved, and even sent to the distribution network in turn to relieve the pressure on the power supply of the grid. Therefore, the net load forecast of households with photovoltaic equipment can help to improve the distribution of electricity in the distribution network. The safety and stability of the power supply and the quality of power supply will help the power grid to allocate power resources according to the power consumption needs of different users and achieve a reasonable allocation of resources.

In order to improve the above problems, the net load prediction method and device of the photovoltaic equipment provided by the present invention include: using a pre-built regression model to predict the sparse solution of the wavelet data of the net load of the photovoltaic equipment in a high-dimensional feature space; Perform wavelet reconstruction of the wavelet data of the net load in the sparse solution of the high-dimensional feature space to obtain the net load prediction data of the photovoltaic equipment; wherein, the wavelet data includes: low-frequency coefficients and high-frequency coefficients. The technical solution provided by the present invention can effectively realize data dimensionality reduction and high-precision prediction of photovoltaic load, and can realize analysis of various external characteristics and power consumption behavior, realize high-dimensional data variable screening and high-precision prediction, and thus rationally arrange The operation plan of the power system;

Further, the technical solution provided by the present invention introduces a prediction model combining wavelet transform with Lasso regression model, in which wavelet transform reverses the time-frequency domain of time series data and focuses on the details of the data, which is more suitable for describing the photovoltaic net load Intrinsic characteristics, and the kernel method is introduced in the Lasso regression model to map the original data to a suitable high-dimensional feature space, so that the Lasso regression model can be applied to the nonlinear photovoltaic net load data, and finally realize the accurate prediction of photovoltaic load and maintain the stability of the large power grid. Stable operation. The above scheme will be described in detail below.

Example 1

Referring to accompanying drawing 1, Fig. 1 is a schematic flowchart of main steps of a method for forecasting net load of photovoltaic equipment according to an embodiment of the present invention. As shown in Figure 1, the net load forecasting method of the photovoltaic equipment in the embodiment of the present invention mainly comprises the following steps:

Step S101: using the pre-built regression model to predict the sparse solution of the wavelet data of the net load of the photovoltaic equipment in the high-dimensional feature space;

Step S102: performing wavelet reconstruction on the sparse solution of the wavelet data of the net load of the photovoltaic equipment in the high-dimensional feature space to obtain the net load prediction data of the photovoltaic equipment;

Wherein, the wavelet data includes: low-frequency coefficients and high-frequency coefficients.

Wavelet transform mainly includes continuous wavelet transform and discrete wavelet transform. Since in this embodiment the net load at the current time is predicted using the data at the same time in the past, this embodiment uses discrete wavelet transform to decompose the original net load time series data.

Decompose the input photovoltaic net load time series, after decomposition, the original signal will produce an approximate series and several detail series. The approximate series represent the low-frequency coefficients (approximate coefficients), which contain the trend information of the photovoltaic net load; the detail series represent the high-frequency coefficients (detail coefficients), which contain the characteristics of the influencing factors related to the photovoltaic net load. Wherein, the calculation formula of the low frequency coefficient of the net load of the photovoltaic equipment is as follows:

The calculation formula of the high frequency coefficient of the net load of the photovoltaic equipment is as follows:

为

对应的分解系数，

为对应缩放常数m和平移常数n选择的小波函数，t为当前时刻，D为光伏设备的净负荷的高频系数，

为ψ_mn对应的分解系数，ψ_mn为与

的互补的小波函数。In the above formula, A is the low frequency coefficient of the net load of the photovoltaic equipment,

for

The corresponding decomposition coefficient,

is the wavelet function selected corresponding to the scaling constant m and the translation constant n, t is the current moment, D is the high frequency coefficient of the net load of the photovoltaic equipment,

is the decomposition coefficient corresponding to ψ _mn , and ψ _mn is the

The complementary wavelet function of .

对应的分解系数的计算式如下：Further, the

The calculation formula of the corresponding decomposition coefficient is as follows:

The calculation formula of the decomposition coefficient corresponding to the ψ _mn is as follows:

In the above formula, T is the net load sequence length of photovoltaic equipment, p _t is the net load of photovoltaic equipment at time t.

Since the application effect of Daubechies (db) is better in time series processing, Daubechies wavelet transform is selected in this embodiment to decompose the original payload data. After the Lasso regression model is predicted, wavelet reconstruction is required to transform the coefficient data Restore the original payload data.

In this embodiment, the low-frequency coefficient and high-frequency coefficient of the photovoltaic net load obtained by wavelet transformation are predicted by using the linear regression method of kernel Lasso. Compared with ordinary linear regression, Lasso can enhance the fitting effect of low-frequency coefficients and high-frequency coefficients of photovoltaic net load, and the use of kernel functions can extend Lasso to non-linear photovoltaic net load data.

The acquisition process of the pre-built regression model includes:

Perform wavelet transformation on the historical net load data of the photovoltaic equipment to obtain the wavelet data of the historical net load data of the photovoltaic equipment;

The wavelet data of the historical net load data of the photovoltaic equipment is extended to the high-dimensional feature space by using the nuclear method, and the linear data corresponding to the wavelet data of the historical net load data of the photovoltaic equipment are obtained;

Using the linear data to construct training data and verification data;

Utilize the training data to train the initial Lasso regression model, utilize the verification data to verify the trained Lasso regression model, until the model index of the trained Lasso regression model meets the convergence condition, obtain the regression model constructed in advance .

Furthermore, the traditional regression method selects the function based on the fact that the sum of squares (ie, variance) of the difference between the low-frequency coefficients obtained by the function through the independent variable and the actual value is the smallest, but this method is prone to over-fitting. In this embodiment, a penalty function is introduced on the basis of the standard loss function, and a shrinkage parameter is multiplied before the absolute value of the coefficient. In this way, the value of the coefficient is shrunk to reduce the variance. The initial regression model using the training data During the training process, the calculation formula of the regression model loss function is as follows:

L ^Reg (β) = L ^OLS (β) + P

In the above formula, L ^Reg is the penalty loss function, L ^OLS is the standard loss function, P is the penalty function value, and β is the regression coefficient vector.

Wherein, the calculation formula of the standard loss function is as follows:

LOLS (β)＝||Y- ^Xβ || ²

In the above formula, Y is an x×1-dimensional predictor variable matrix, X is an x×p-dimensional outcome variable vector, x is the number of observations, and p is the number of predictors.

Using the Lasso model to predict the low and high frequency coefficients of the photovoltaic net load obtained by wavelet transform can effectively prevent overfitting. In addition, whether it is the case of relatively small training data or the case of too high dimension, continuous or discrete, Lasso can handle it all.

On the other hand, the low and high frequency coefficients of photovoltaic net load obtained by wavelet transform may be nonlinear, and the conventional Lasso method cannot handle nonlinear data. For these data types that cannot be processed by linear regression, the original data can be extended to High-dimensional space, perform linear regression in high-dimensional space.

Kernel methods (KMs) are a class of algorithms for pattern recognition. Its purpose is to find and learn the mutual relationship in a set of data. Widely used kernel methods include support vector machines, Gaussian processes, etc.

Through the kernel method, the nonlinear low- and high-frequency coefficients of PV net loads can be embedded into a suitable high-dimensional feature space; then, patterns are analyzed and processed in this new space using a general linear learner. The existence of the kernel function (Kernel function) can imply the nonlinear mapping in the linear learner for simultaneous calculation, so that the computational complexity has nothing to do with the dimension of the high-dimensional feature space.

The data of low and high frequency coefficients of photovoltaic net load is extended to the data of high-dimensional feature space by kernel function for Lasso fitting, so as to obtain the sparse solution of high-dimensional feature space.

In an optimal implementation, the kernel Lasso regression prediction is mainly performed for each dimension of the 96-dimensional net load data. Before that, it is necessary to perform discrete wavelet transform on each dimension of the photovoltaic net load data to obtain high-frequency coefficients D and low-frequency coefficient A, then train and predict D and A in each dimension respectively to obtain predicted D and A, then perform wavelet reconstruction to obtain a predicted value of one dimension, and then combine the predicted values of 96 dimensions It is the forecast of one day, and then compare it with the actual value to analyze the forecast effect.

In this embodiment, the model index includes at least one of the following: mean absolute percentage error, mean square error.

In one embodiment, the formula for calculating the mean absolute percentage error is as follows:

The formula for calculating the mean square error is as follows:

为所述验证数据中第i个样本数据的预测值，MAPE为均方误差。In the above formula, MSE is the mean absolute percentage error, I is the total number of sample data in the verification data, and y _i is the actual value of the i-th sample data in the verification data,

is the predicted value of the i-th sample data in the verification data, and MAPE is the mean square error.

In an optimal implementation, the net load forecasting method of photovoltaic equipment provided by the present invention is used for forecasting, and the overall forecasting effect is good. Here is a weekly forecast for a certain station area for 2021/5/12-2021/5/18 Indicators, as shown in Table 1:

Table 1

It can be seen from this that the prediction effect of the net load prediction method for photovoltaic equipment provided by the present invention is good.

Then choose the wavelet transform combined with the kernel Lasso regression algorithm to predict each station area, use the historical data before the week to be predicted as the training set, and make weekly predictions for the week after training, from 2021/1/1-2021/1/7 The results of the weekly forecast index values are shown in Table 2:

Table 2

It can be seen from this that the kernel Lasso regression algorithm has a better prediction effect on most of the station areas, but the effect on station area 5 and station area 9 is very poor. After analysis, the predicted and actual values of station area 5 and station area 9 The difference is mainly reflected in the peak value of the photovoltaic net load. The large difference between the peak value and its vicinity will eventually lead to the prediction effect. Although the overall prediction effect of other stations is better, the prediction effect near the peak value is not ideal. The possible reasons There may be two reasons. One is that Kernel Lasso’s peak prediction effect is very poor and cannot quickly reflect changes in time series. The other reason is that it does not consider the influence of meteorological factors such as temperature and light on the payload data.

Further, the effect of using wavelet transform

Finally, we can compare using only the kernel Lasso regression model and using the wavelet transform combined with the kernel Lasso regression model to make predictions. It is found that in some data areas, the accuracy of the latter is much higher than that of the former. It is shown below in a certain area in 2021/5/ The results of the weekly forecast indicators for 5-2021/5/11 are shown in Table 3:

The reason why the effect is good is mainly because the wavelet transform can divide the time series into more stable and smooth coefficient series, which is convenient for the prediction of the subsequent kernel Lasso regression model. Therefore, it is possible to combine the kernel Lasso regression model based on wavelet transform with the kernel Lasso regression model only, that is, to select a model with a smaller index value for prediction according to the index value of the last prediction, and continue to improve the overall prediction. accuracy.

Example 2

Based on the same inventive concept, the present invention also provides a net load forecasting device for photovoltaic equipment, as shown in Figure 2, the net load forecasting device for photovoltaic equipment includes:

The prediction module is used to predict the sparse solution of the wavelet data of the net load of the photovoltaic equipment in the high-dimensional feature space by using the regression model constructed in advance;

The reconstruction module is used to perform wavelet reconstruction on the sparse solution of the wavelet data of the net load of the photovoltaic equipment in the high-dimensional feature space, and obtain the net load prediction data of the photovoltaic equipment;

Wherein, the wavelet data includes: low-frequency coefficients and high-frequency coefficients.

Preferably, in the prediction module, the calculation formula of the low frequency coefficient of the net load of the photovoltaic equipment is as follows:

The calculation formula of the high frequency coefficient of the net load of the photovoltaic equipment is as follows:

为

对应的分解系数，

为对应缩放常数m和平移常数n选择的小波函数，t为当前时刻，D为光伏设备的净负荷的高频系数，

为ψ_mn对应的分解系数，ψ_mn为与

的互补的小波函数。In the above formula, A is the low frequency coefficient of the net load of the photovoltaic equipment,

for

The corresponding decomposition coefficient,

is the wavelet function selected corresponding to the scaling constant m and the translation constant n, t is the current moment, D is the high frequency coefficient of the net load of the photovoltaic equipment,

is the decomposition coefficient corresponding to ψ _mn , and ψ _mn is the

The complementary wavelet function of .

对应的分解系数的计算式如下：Further, the

The calculation formula of the corresponding decomposition coefficient is as follows:

The calculation formula of the decomposition coefficient corresponding to the ψ _mn is as follows:

In the above formula, T is the net load sequence length of photovoltaic equipment, p _t is the net load of photovoltaic equipment at time t.

Preferably, in the prediction module, the acquisition process of the pre-built regression model includes:

Perform wavelet transformation on the historical net load data of the photovoltaic equipment to obtain the wavelet data of the historical net load data of the photovoltaic equipment;

Using nuclear devices to expand the wavelet data of the historical net load data of photovoltaic equipment to high-dimensional feature space, and obtain the linear data corresponding to the wavelet data of the historical net load data of photovoltaic equipment;

Using the linear data to construct training data and verification data;

Utilize the training data to train the initial Lasso regression model, utilize the verification data to verify the trained Lasso regression model, until the model index of the trained Lasso regression model meets the convergence condition, obtain the regression model constructed in advance .

Further, in the process of using the training data to train the initial regression model, the calculation formula of the regression model loss function is as follows:

L ^Reg (β) = L ^OLS (β) + P

In the above formula, L ^Reg is the penalty loss function, LOLS is the standard loss function, P is the penalty function value, and β is the regression coefficient vector.

Further, the calculation formula of the standard loss function is as follows:

LOLS (β)＝||Y- ^Xβ || ²

In the above formula, Y is an x×1-dimensional predictor variable matrix, X is an x×p-dimensional outcome variable vector, x is the number of observations, and p is the number of predictors.

Further, the model index includes at least one of the following: mean absolute percentage error, mean square error.

Further, the formula for calculating the mean absolute percentage error is as follows:

The formula for calculating the mean square error is as follows:

为所述验证数据中第i个样本数据的预测值，MAPE为均方误差。In the above formula, MSE is the mean absolute percentage error, I is the total number of sample data in the verification data, and y _i is the actual value of the i-th sample data in the verification data,

is the predicted value of the i-th sample data in the verification data, and MAPE is the mean square error.

Example 3

Based on the same inventive concept, the present invention also provides a computer device, the computer device includes a processor and a memory, the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the Program instructions stored in a computer storage medium. The processor may be a central processing unit (Central Processing Unit, CPU), or other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), off-the-shelf programmable gate arrays (Field -Programmable GateArray, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, which are suitable for implementing one or more instructions, and are specifically suitable for loading and Execute one or more instructions in the computer storage medium to realize the corresponding method flow or corresponding function, so as to realize the steps of a method for predicting the net load of photovoltaic equipment in the above embodiment.

Example 4

Based on the same inventive concept, the present invention also provides a storage medium, specifically a computer-readable storage medium (Memory). The computer-readable storage medium is a memory device in a computer device for storing programs and data. It can be understood that the computer-readable storage medium here may include a built-in storage medium in the computer device, and of course may also include an extended storage medium supported by the computer device. The computer-readable storage medium provides storage space, and the storage space stores the operating system of the terminal. Moreover, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions may be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here may be a high-speed RAM memory, or a non-volatile memory (non-volatile memory), such as at least one magnetic disk memory. One or more instructions stored in the computer-readable storage medium can be loaded and executed by the processor, so as to realize the steps of a method for predicting the net load of photovoltaic equipment in the above-mentioned embodiments.

Those skilled in the art should understand that the embodiments of the present invention may be provided as methods, systems, or computer program products. Accordingly, the present invention can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each procedure and/or block in the flowchart and/or block diagram, and a combination of procedures and/or blocks in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions may be provided to a general purpose computer, special purpose computer, embedded processor, or processor of other programmable data processing equipment to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing equipment produce a An apparatus for realizing the functions specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to operate in a specific manner, such that the instructions stored in the computer-readable memory produce an article of manufacture comprising instruction means, the instructions The device realizes the function specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions can also be loaded onto a computer or other programmable data processing device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby The instructions provide steps for implementing the functions specified in the flow chart or blocks of the flowchart and/or the block or blocks of the block diagrams.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: the present invention can still be Any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention shall fall within the protection scope of the claims of the present invention.

Claims (18)

1. A net load forecasting method for photovoltaic equipment, characterized in that the method comprises: Sparse solution of wavelet data in high-dimensional feature space to predict net load of photovoltaic equipment using pre-built regression model; Performing wavelet reconstruction on the sparse solution of the wavelet data of the net load of the photovoltaic equipment in the high-dimensional feature space, to obtain the net load prediction data of the photovoltaic equipment; Wherein, the wavelet data includes: low-frequency coefficients and high-frequency coefficients. 2. The method according to claim 1, wherein the calculation formula of the low-frequency coefficient of the net load of the photovoltaic device is as follows:

The calculation formula of the high frequency coefficient of the net load of the photovoltaic equipment is as follows:

In the above formula, A is the low frequency coefficient of the net load of the photovoltaic equipment,

for

The corresponding decomposition coefficient,

is the wavelet function selected corresponding to the scaling constant m and the translation constant n, t is the current moment, D is the high frequency coefficient of the net load of the photovoltaic equipment,

is the decomposition coefficient corresponding to ψ _mn , and ψ _mn is the

The complementary wavelet function of . 3. The method of claim 2, wherein the

The calculation formula of the corresponding decomposition coefficient is as follows:

The calculation formula of the decomposition coefficient corresponding to the ψ _mn is as follows:

In the above formula, T is the net load sequence length of photovoltaic equipment, p _t is the net load of photovoltaic equipment at time t. 4. The method according to claim 1, wherein the acquisition process of the pre-built regression model comprises: Perform wavelet transformation on the historical net load data of the photovoltaic equipment to obtain the wavelet data of the historical net load data of the photovoltaic equipment; The wavelet data of the historical net load data of the photovoltaic equipment is extended to the high-dimensional feature space by using the nuclear method, and the linear data corresponding to the wavelet data of the historical net load data of the photovoltaic equipment are obtained; Using the linear data to construct training data and verification data; Utilize the training data to train the initial Lasso regression model, utilize the verification data to verify the trained Lasso regression model, until the model index of the trained Lasso regression model meets the convergence condition, obtain the regression model constructed in advance . 5. The method according to claim 4, wherein, in the process of utilizing the training data to train the initial regression model, the calculation formula of the regression model loss function is as follows: L ^Reg (β) = L ^OLS (β) + P In the above formula, L ^Reg is the penalty loss function, L ^OLS is the standard loss function, P is the penalty function value, and β is the regression coefficient vector. 6. The method according to claim 5, wherein the calculation formula of the standard loss function is as follows: LOLS (β)＝||Y- ^Xβ || ² In the above formula, Y is an x×1-dimensional predictor variable matrix, X is an x×p-dimensional outcome variable vector, x is the number of observations, and p is the number of predictors. 7. The method according to claim 4, wherein the model index comprises at least one of the following: mean absolute percentage error, mean square error. 8. method as claimed in claim 7, is characterized in that, the computing formula of described mean absolute percentage error is as follows:

The formula for calculating the mean square error is as follows:

In the above formula, MSE is the mean absolute percentage error, I is the total number of sample data in the verification data, and y _i is the actual value of the i-th sample data in the verification data,

is the predicted value of the i-th sample data in the verification data, and MAPE is the mean square error. 9. A net load forecasting device for photovoltaic equipment, characterized in that the device comprises: The prediction module is used to predict the sparse solution of the wavelet data of the net load of the photovoltaic equipment in the high-dimensional feature space by using the regression model constructed in advance; The reconstruction module is used to perform wavelet reconstruction on the sparse solution of the wavelet data of the net load of the photovoltaic equipment in the high-dimensional feature space, and obtain the net load prediction data of the photovoltaic equipment; Wherein, the wavelet data includes: low-frequency coefficients and high-frequency coefficients. 10. The device according to claim 9, wherein, in the prediction module, the calculation formula of the low-frequency coefficient of the net load of photovoltaic equipment is as follows:

The calculation formula of the high frequency coefficient of the net load of the photovoltaic equipment is as follows:

In the above formula, A is the low frequency coefficient of the net load of the photovoltaic equipment,

for

The corresponding decomposition coefficient,

is the wavelet function selected corresponding to the scaling constant m and the translation constant n, t is the current moment, D is the high frequency coefficient of the net load of the photovoltaic equipment,

is the decomposition coefficient corresponding to ψ _mn , and ψ _mn is the

The complementary wavelet function of . 11. The apparatus of claim 10, wherein the

The calculation formula of the corresponding decomposition coefficient is as follows:

The calculation formula of the decomposition coefficient corresponding to the ψ _mn is as follows:

In the above formula, T is the net load sequence length of photovoltaic equipment, p _t is the net load of photovoltaic equipment at time t. 12. The device according to claim 9, wherein, in the prediction module, the acquisition process of the pre-built regression model comprises: Perform wavelet transformation on the historical net load data of the photovoltaic equipment to obtain the wavelet data of the historical net load data of the photovoltaic equipment; Using nuclear devices to expand the wavelet data of the historical net load data of photovoltaic equipment to high-dimensional feature space, and obtain the linear data corresponding to the wavelet data of the historical net load data of photovoltaic equipment; Using the linear data to construct training data and verification data; Utilize the training data to train the initial Lasso regression model, utilize the verification data to verify the trained Lasso regression model, until the model index of the trained Lasso regression model meets the convergence condition, obtain the regression model constructed in advance . 13. The device according to claim 12, wherein, in the process of using the training data to train the initial regression model, the calculation formula of the regression model loss function is as follows: L ^Reg (β) = L ^OLS (β) + P In the above formula, L ^Reg is the penalty loss function, L ^OLS is the standard loss function, P is the penalty function value, and β is the regression coefficient vector. 14. The device according to claim 13, wherein the calculation formula of the standard loss function is as follows: LOLS (β)＝||Y- ^Xβ || ² In the above formula, Y is an x×1-dimensional predictor variable matrix, X is an x×p-dimensional outcome variable vector, x is the number of observations, and p is the number of predictors. 15. The device according to claim 12, wherein the model index comprises at least one of the following: mean absolute percentage error, mean square error. 16. The device according to claim 15, wherein the formula for calculating the mean absolute percentage error is as follows:

The formula for calculating the mean square error is as follows:

In the above formula, MSE is the mean absolute percentage error, I is the total number of sample data in the verification data, and y _i is the actual value of the i-th sample data in the verification data,

is the predicted value of the i-th sample data in the verification data, and MAPE is the mean square error. 17. A computer device, comprising: one or more processors; The processor is configured to store one or more programs; When the one or more programs are executed by the one or more processors, the net load forecasting method for a photovoltaic device according to any one of claims 1 to 8 is realized. 18. A computer-readable storage medium, characterized in that there is a computer program stored thereon, and when the computer program is executed, the net load prediction method of photovoltaic equipment as claimed in any one of claims 1 to 8 is realized .

Images