Regional renewable energy source configuration method and system based on electric power carbon emission

Abstract

The invention discloses a regional renewable energy source configuration method and system based on electric power carbon emission, and belongs to the technical field of carbon emission, and the method comprises the steps: S1, collecting the data of regional historical power generation, load power and the like, and building a regional source-charge unbalance prediction model; s2, constructing a renewable energy investment model by taking the lowest total cost of regional power generation as a target, and determining the renewable energy investment capacity; s3, based on the power flow information, establishing a regional node power carbon emission factor library; and S4, determining the position of the power grid node where the renewable energy project is located by taking the minimum regional carbon emission as a target and combining the power consumption of the regional power grid node and the power carbon emission factor. According to the method, the overall operation cost of the regional power system can be effectively reduced, the power grid carbon emission characteristics are considered, renewable energy sources are configured by identifying the high-carbon emission nodes in the regional power grid, and the total regional power carbon emission amount can be effectively reduced.

Description

Technical Field

[0001] The present invention relates to a low-carbon configuration of distributed power sources, and in particular to a method and system for configuring regional renewable energy based on power carbon emissions. Background Art

[0002] With the continuous increase in global energy demand and the continuous improvement of environmental protection awareness, the consumption of renewable energy such as solar energy and wind energy continues to grow. The efficient and economical configuration of renewable energy will continue to help improve the utilization efficiency of regional renewable energy, reduce carbon emissions in power production, and assist in regional energy management, promote grid structure optimization and green and low-carbon transformation.

[0003] Existing regional renewable energy configuration methods often carry out modeling with the goal of optimizing their own development, but fail to effectively combine the carbon emission characteristics of regional power grids for development and layout, weakening the supporting role of distributed renewable energy development in optimizing the overall regional energy structure and green and low-carbon development, and failing to maximize carbon reduction benefits at the regional macro-carbon reduction level. Summary of the invention

[0004] The purpose of the present invention is to address the deficiencies of the above-mentioned background technology and to provide a method and system for regional renewable energy configuration based on electricity carbon emissions.

[0005] To achieve the above objectives, the technical solution of the present invention is:

[0006] In a first aspect, the present invention proposes a method for configuring regional renewable energy based on electricity carbon emissions, comprising:

[0007] S1. Collect historical power generation and load data of the target area, establish a regional source-load imbalance prediction model, and determine the target area source-load imbalance prediction value;

[0008] S2. Determine the newly added renewable energy installed capacity based on the predicted value of source-load imbalance in the target area;

[0009] S3. Establish a regional node power carbon emission factor library based on power flow information;

[0010] S4. Based on the newly added installed capacity of renewable energy, with the goal of minimizing regional carbon emissions, combined with the regional grid node power consumption and electricity carbon emission factors, determine the grid node location of the renewable energy project.

[0011] In S1, the regional source-charge imbalance prediction model constructed in 1.1 is:

[0012] Q balance,t =Q source,t -Q load,t

[0013] In the formula, Q balance,t , Q source,t , Q load,t They are the unbalanced power, generated power and load power in the target area during period t respectively.

[0014] 1.2 Power generation Q source,t The predicted values ​​are:

[0015]

[0016] Where K is the number of regional power types, Q k,t is the predicted value of the power generation of the kth power source in the region during period t.

[0017] The values ​​of the Spearman correlation coefficient are:

[0018]

[0019] Where ρ is the Spearman correlation coefficient, R(x j )、R(y j ) are the positions of variables x and y respectively, are the average ranks of variables x and y respectively, and m is the number of samples.

[0020] The Stacking algorithm is used to calculate the power generation Q of different power types in the target area. k,t To make a prediction, the prediction steps are:

[0021] 1.2.1 Use the Spearman correlation coefficient method to analyze the correlation between power generation influencing factors and power generation of different types of power sources. Rank the power generation influencing factors based on the Spearman correlation coefficient, select the historical data of power generation influencing factors with the top n correlation rankings, and combine them with the historical power generation to form the initial data set of different power types. Perform steps 1.2.2-1.2.4 for each power type;

[0022] 1.2.2 Divide the initial data set into a meta-training set and a meta-test set on the meta-learner, and divide the meta-training set into a base training set and a base test set on the base learner;

[0023] 1.2.3 Train and predict the base learners. Select ResNet, VGG, and MobileNet algorithm models as base learners. Use K-fold cross-validation to train the model on K-1 folds and verify it on the Kth fold. Each base learner is trained K times, for a total of 3K times, to generate all base models and prediction results.

[0024] 1.2.4 Based on the prediction data of the base learner, a prediction model is made on the meta-learner. Linear regression and logistic regression algorithms are used as meta-learners respectively. The training results of the two learners are averaged to obtain the final prediction result.

[0025] 1.3 Regional power load forecasting is carried out through the multi-layer perceptron model based on empirical mode decomposition. The specific steps are:

[0026] 1.3.1 Construct a regional historical load power data set, divide the regional historical load power data set into a training set and a test set, and use the empirical mode decomposition algorithm to decompose the historical load power data into n intrinsic mode components and 1 residual:

[0027]

[0028] In the formula, Q load 、IMF i , Res are the historical load power, the i-th intrinsic mode component, and the residual respectively;

[0029] 1.3.2 The decomposed intrinsic mode components are reconstructed using the extreme point partitioning method to form high-frequency, medium-frequency and low-frequency components. A multi-layer perceptron prediction model is established for each frequency band, namely HF-MLP, IF-MLP and LF-MLP;

[0030] 1.3.3 The Spearman correlation coefficient method is used to analyze the correlation between candidate features and high-frequency, medium-frequency, and low-frequency components, and the components with absolute values ​​of correlation coefficients greater than the set threshold are selected as the inputs of the HF-MLP, IF-MLP, and LF-MLP models;

[0031] 1.3.4 Use the training set to train and adjust the parameters of the HF-MLP, IF-MLP, and LF-MLP models. After the model is trained, the high-frequency, medium-frequency, and low-frequency components of the power load data are predicted on the test set, and the prediction results are superimposed to obtain the load power prediction value Q load,t .

[0032] In S2: based on the target area source-charge unbalanced power prediction value Q balance,t , considering the cost of new renewable energy power generation P invest,t , the cost of electricity generation from existing renewable energy sources P genera,t 、The cost of electricity transferred from other regions P buy,t And converted into green certificate income P green,t , with the total regional power generation cost P all,t Minimum goal: building a renewable energy investment model:

[0033] minP all,t =min(P invest,t +Pgenera,t +P buy,t -P green,t )

[0034]

[0035] P buy,t =-Q balance,t ×W t

[0036]

[0037] Where R is the type of new renewable energy source; G r,t , U r,t , H r,t , r,t are the installed capacity, unit installed capacity cost, power generation utilization hours, and unit power operation and maintenance cost coefficient of the rth type of newly added renewable energy power source in period t; b is the discount rate; T r Add the service life of renewable energy source for category r; V k,t W is the electricity cost of the kth type of stock power source in the region during period t; t is the inter-regional electricity transaction unit price during period t; S is the type of renewable energy in the stock power source; Q s,t is the power generation of the sth stock of renewable energy in period t; δ is the quantitative coefficient of renewable energy power generation converted into green certificates, I t It is the average transaction price of green certificates during period t.

[0038] In S3, the node power carbon emission factor is obtained based on the power flow, which is the average carbon emission factor of node a in the target area during period t, and is:

[0039]

[0040] In the formula, c a is the average grid carbon emission factor of regional node a, is the direct carbon emission of the power plant directly connected to node a, is the indirect carbon emission injected into the carbon emission flow of node a, Q a is the load power of node a, Q out is the power flow out of node a; the target area has A nodes, and the power carbon emission factor library has {c1, c2, ..., c A}.

[0041] In S4: according to the newly added renewable energy installed capacity, the newly added renewable energy power source node is placed at node x, and a regional power carbon emission model is established, which is:

[0042]

[0043] In the formula, Γ area,t is the total electricity carbon emissions of the region in period t, c a,t , Q a,t are the average grid carbon emission factor and load power of node a in period t, x is the node where the newly added renewable energy source is located, and 1≤x≤A, c x,t is the average grid carbon emission factor of node x during period t;

[0044] The newly added renewable energy power sources are placed at Node 1, Node 2, ..., Node A respectively, and the regional electricity carbon emissions under each scenario are calculated. The scenario with the smallest regional electricity carbon emissions is taken as the grid node location where the newly added renewable energy power source is located.

[0045] In a second aspect, the present invention proposes a regional renewable energy configuration system based on electricity carbon emissions, comprising:

[0046] Regional source-charge imbalance prediction model building module: used to collect historical power generation and load data of the target area and establish a regional source-charge imbalance prediction model;

[0047] Renewable energy investment model building module: used to build a renewable energy investment model with the goal of minimizing the total cost of regional power generation and determine the newly added renewable energy installed capacity;

[0048] Regional node power carbon emission factor library construction module: used to establish a regional node power carbon emission factor library based on power flow information;

[0049] Renewable energy configuration plan generation module: It is used to determine the grid node location of the renewable energy project with the goal of minimizing regional carbon emissions, combined with the regional grid node power consumption and electricity carbon emission factors.

[0050] In S1, the regional source-charge imbalance prediction model constructed in 1.1 is:

[0051] Q balance,t =Q source,t -Q load,t

[0052] In the formula, Q balance,t , Q source,t , Q load,t They are the unbalanced power, generated power and load power in the target area during period t respectively.

[0053] 1.2 Power generation Q source,t The predicted values ​​are:

[0054]

[0055] Where K is the number of regional power types, Q k,t is the predicted value of the power generation of the kth power source in the region during period t.

[0056] The values ​​of the Spearman correlation coefficient are:

[0057]

[0058] Where ρ is the Spearman correlation coefficient, R(x j )、R(y j ) are the positions of variables x and y respectively, are the average ranks of variables x and y respectively, and m is the number of samples.

[0059] The Stacking algorithm is used to calculate the power generation Q of different power types in the target area. k,t To make a prediction, the prediction steps are:

[0060] 1.2.1 Use the Spearman correlation coefficient method to analyze the correlation between power generation influencing factors and power generation of different types of power sources. Rank the power generation influencing factors based on the Spearman correlation coefficient, select the historical data of power generation influencing factors with the top n correlation rankings, and combine them with the historical power generation to form the initial data set of different power types. Perform steps 1.2.2-1.2.4 for each power type;

[0061] 1.2.2 Divide the initial data set into a meta-training set and a meta-test set on the meta-learner, and divide the meta-training set into a base training set and a base test set on the base learner;

[0062] 1.2.3 Train and predict the base learners. Select ResNet, VGG, and MobileNet algorithm models as base learners. Use K-fold cross-validation to train the model on K-1 folds and verify it on the Kth fold. Each base learner is trained K times, for a total of 3K times, to generate all base models and prediction results.

[0063] 1.2.4 Based on the prediction data of the base learner, a prediction model is made on the meta-learner. Linear regression and logistic regression algorithms are used as meta-learners respectively. The training results of the two learners are averaged to obtain the final prediction result.

[0064] 1.3 Regional power load forecasting is carried out through the multi-layer perceptron model based on empirical mode decomposition. The specific steps are:

[0065] 1.3.1 Construct a regional historical load power data set, divide the regional historical load power data set into a training set and a test set, and use the empirical mode decomposition algorithm to decompose the historical load power data into n intrinsic mode components and 1 residual:

[0066]

[0067] In the formula, Q load 、IMF i , Res are the historical load power, the i-th intrinsic mode component, and the residual respectively;

[0068] 1.3.2 The decomposed intrinsic mode components are reconstructed using the extreme point partitioning method to form high-frequency, medium-frequency and low-frequency components. A multi-layer perceptron prediction model is established for each frequency band, namely HF-MLP, IF-MLP and LF-MLP;

[0069] 1.3.3 The Spearman correlation coefficient method is used to analyze the correlation between candidate features and high-frequency, medium-frequency, and low-frequency components, and the components with absolute values ​​of correlation coefficients greater than the set threshold are selected as the inputs of the HF-MLP, IF-MLP, and LF-MLP models;

[0070] 1.3.4 Use the training set to train and adjust the parameters of the HF-MLP, IF-MLP, and LF-MLP models. After the model is trained, the high-frequency, medium-frequency, and low-frequency components of the power load data are predicted on the test set, and the prediction results are superimposed to obtain the load power prediction value Q load,t .

[0071] In S2: Consider the cost of new renewable energy power generation P invest,t , the cost of electricity generation from existing renewable energy sources P genera,t 、The cost of electricity transferred from other regions P buy,t And converted into green certificate income P green,t , with the total regional power generation cost P all,t Minimum goal: building a renewable energy investment model:

[0072] minP all,t =min(P invest,t +P genera,t +P buy,t -P green,t )

[0073]

[0074] P buy,t =-Q balance,t ×W t

[0075]

[0076] Where R is the type of new renewable energy source; G r,t , U r,t , H r,t , r,tare the installed capacity, unit installed capacity cost, power generation utilization hours, and unit power operation and maintenance cost coefficient of the rth type of newly added renewable energy power source in period t; b is the discount rate; T r Add the service life of renewable energy source for category r; V k,t W is the electricity cost of the kth type of stock power source in the region during period t; t is the inter-regional electricity transaction unit price during period t; S is the type of renewable energy in the stock power source; Q s,t is the power generation of the sth stock of renewable energy in period t; δ is the quantitative coefficient of renewable energy power generation converted into green certificates, I t It is the average transaction price of green certificates during period t.

[0077] In S3, the node power carbon emission factor is obtained based on the power flow, which is the average carbon emission factor of node a in the target area during period t, and is:

[0078]

[0079] In the formula, c a is the average grid carbon emission factor of regional node a, is the direct carbon emission of the power plant directly connected to node a, is the indirect carbon emission injected into the carbon emission flow of node a, Q a is the load power of node a, Q out is the power flow out of node a; the target area has A nodes, and the power carbon emission factor library has {c1, c2, ..., c A}.

[0080] In S4: according to the newly added renewable energy installed capacity, the newly added renewable energy power source node is placed at node x, and a regional power carbon emission model is established, which is:

[0081]

[0082] In the formula, Γ area,t is the total electricity carbon emissions of the region in period t, c a,t , Q a,t are the average grid carbon emission factor and load power of node a in period t, x is the node where the newly added renewable energy source is located, and 1≤x≤A, c x,t is the average grid carbon emission factor of node x during period t;

[0083] The newly added renewable energy power sources are placed at Node 1, Node 2, ..., Node A respectively, and the regional electricity carbon emissions under each scenario are calculated. The scenario with the smallest regional electricity carbon emissions is taken as the grid node location where the newly added renewable energy power source is located.

[0084] Compared with the prior art, the present invention has the following beneficial effects:

[0085] The present invention is a regional source-charge imbalance prediction model constructed in a regional renewable energy configuration method based on electricity carbon emissions. On the one hand, the Stacking algorithm is used to predict the power generation of different power types separately, and then the multi-layer perceptron (EMD-MLP) model based on empirical mode decomposition is used to carry out regional power load prediction. Finally, the results of the above two predictions are used to calculate the predicted value of the regional source-charge imbalance. The prediction result is more accurate and the prediction accuracy is higher. On the other hand, a renewable energy investment model, a regional node electricity carbon emission factor library, and a regional electricity carbon emission model are constructed. According to the investment capacity calculated by the renewable energy investment model, calculations are performed one by one in the regional electricity carbon emission model, and the scenario with the smallest regional electricity carbon emission is selected as the grid node location where the new renewable energy power source is located. While reducing the overall operating cost of the regional power system, this design can effectively reduce the total amount of regional electricity carbon emissions and help the green and low-carbon transformation of the regional power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 It is a flow chart of the method of the present invention.

[0087] Figure 2 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0088] The present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0089] Embodiment 1:

[0090] See also Figure 1 , a regional renewable energy configuration method based on electricity carbon emissions, comprising:

[0091] S1. Collect historical power generation and load data of the target area and establish a regional source-load imbalance prediction model;

[0092] 1.1 The regional source-charge imbalance prediction model constructed is:

[0093] Q balance,t =Q source,t -Q load,t

[0094] In the formula, Q balance,t、 Q source,t、 Q load,t They are the unbalanced power, generated power and load power in the target area during period t respectively.

[0095] 1.2 Power generation Q source,t The predicted values ​​are:

[0096]

[0097] Where K is the number of regional power types, Q k,t is the predicted value of the power generation of the kth power source in the region during period t.

[0098] The values ​​of the Spearman correlation coefficient are:

[0099]

[0100] Where ρ is the Spearman correlation coefficient, R(x j )、R(y j ) are the positions of variables x and y respectively, are the average ranks of variables x and y respectively, and m is the number of samples.

[0101] The Stacking algorithm is used to calculate the power generation Q of different power types in the target area. k,t To make a prediction, the prediction steps are:

[0102] 1.2.1 Use the Spearman correlation coefficient method to analyze the correlation between power generation influencing factors and power generation of different types of power sources. Rank the power generation influencing factors based on the Spearman correlation coefficient, select the historical data of the top two power generation influencing factors, and combine them with the historical power generation to form the initial data set of different power types. Perform steps 1.2.2-1.2.4 for each power type.

[0103] Selecting the historical data of the top two factors affecting power generation with the highest relevance is helpful to reduce the amount of calculation;

[0104] 1.2.2 Divide the initial data set into a meta-training set and a meta-test set on the meta-learner, and divide the meta-training set into a base training set and a base test set on the base learner;

[0105] 1.2.3 Carry out training and prediction for the base learners, select ResNet (Residual Network), VGG (Visual Geometry Group Convolutional Neural Network), MobileNet (Lightweight Convolutional Neural Network) algorithm models as base learners, use K-fold cross-validation method, train the model on K-1 folds, and verify it on the Kth fold. Each base learner is trained K times, for a total of 3K times, to generate all base models and prediction results;

[0106] 1.2.4 Based on the prediction data of the base learner, a prediction model is made on the meta-learner. Linear regression and logistic regression algorithms are used as meta-learners respectively. The training results of the two learners are averaged to obtain the final prediction result.

[0107] After training on the meta-training set, the model is obtained and the meta-test set is used directly for prediction.

[0108] 1.3 Regional power load forecasting is carried out through the multi-layer perceptron model based on empirical mode decomposition. The specific steps are:

[0109] 1.3.1 Construct a regional historical load power data set, divide the regional historical load power data set into a training set and a test set, and use the empirical mode decomposition algorithm to decompose the historical load power data into n intrinsic mode components and 1 residual:

[0110]

[0111] In the formula, Q load 、IMF i , Res are the historical load power, the i-th intrinsic mode component, and the residual respectively;

[0112] 1.3.2 The decomposed intrinsic mode components are reconstructed using the extreme point partitioning method to form high-frequency, medium-frequency and low-frequency components. A multi-layer perceptron prediction model is established for each frequency band, namely HF-MLP, IF-MLP and LF-MLP.

[0113] 1.3.3 The Spearman correlation coefficient method was used to analyze the correlation between the candidate features and the high-frequency, medium-frequency, and low-frequency components, and the components with absolute values ​​of correlation coefficients greater than 0.6 were selected as the inputs of the HF-MLP, IF-MLP, and LF-MLP models.

[0114] Candidate features are factors that can affect load power, such as temperature and humidity. When the absolute value of the correlation coefficient is greater than 0.6, it has a strong correlation.

[0115] 1.3.4 Use the training set to train and adjust the parameters of the HF-MLP, IF-MLP, and LF-MLP models. After the model is trained, the high-frequency, medium-frequency, and low-frequency components of the power load data are predicted on the test set, and the prediction results are superimposed to obtain the load power prediction value Q load,t .

[0116] The predicted value of regional source-charge imbalance is:

[0117] Q balance,t =Q source,t -Q load,t

[0118] In the formula, Q balanc,et , Q source,t , Q load,t They are respectively the unbalanced electricity, generated electricity and load electricity of the region in period t.

[0119] Predicting the high-frequency, medium-frequency, and low-frequency components of power load data separately on the test set can improve the prediction accuracy of the model.

[0120] S2. With the goal of minimizing the total cost of regional power generation, a renewable energy investment model is constructed to determine the newly added renewable energy installed capacity;

[0121] Considering the cost of new renewable energy power generation P invest,t , the cost of electricity generation from existing renewable energy sources P genera,t 、The cost of electricity transferred from other regions P buy,t And converted into green certificate income P green,t , with the total regional power generation cost P all,t Minimum goal: building a renewable energy investment model:

[0122] minP all,t =min(P invest,t +P genera,t +P buy,t -P green,t )

[0123]

[0124] P buy,t =-Q balance,t ×W t

[0125]

[0126] Where R is the type of new renewable energy source; G r,t , U r,t , H r,t , r,t are the installed capacity, unit installed capacity cost, power generation utilization hours, and unit power operation and maintenance cost coefficient of the rth type of newly added renewable energy power source in period t; b is the discount rate; T r Add the service life of renewable energy source for category r; V k,t W is the electricity cost of the kth type of stock power source in the region during period t; t is the inter-regional electricity transaction unit price during period t; S is the type of renewable energy in the stock power source; Q s,t is the power generation of the sth stock of renewable energy in period t; δ is the quantitative coefficient of renewable energy power generation converted into green certificates, I t It is the average transaction price of green certificates during period t.

[0127] The four parameters of the installed capacity of new renewable energy sources, the unit installed capacity cost, the number of hours of power generation, and the unit power operation and maintenance cost coefficient are interrelated. In practice, the installed capacity is determined, and the other parameters can be estimated. Here, it can be regarded as only the installed capacity as an unknown number. It can be regarded as several renewable energy construction plans, which can be calculated separately, and then selected according to the calculation results.

[0128] S3. Establish a regional node power carbon emission factor library based on power flow information;

[0129] The node power carbon emission factor is obtained based on the power flow, which is the average carbon emission factor of node a in the target area during period t, and is:

[0130]

[0131] In the formula, c a is the average grid carbon emission factor of regional node a, is the direct carbon emission of the power plant directly connected to node a, is the indirect carbon emission injected into the carbon emission flow of node a, Q a is the load power of node a, Q out is the power flow out of node a; the target area has A nodes, and the power carbon emission factor library has {c1, c2, ..., c A}.

[0132] S4. Taking the minimum regional carbon emissions as the goal, combined with the regional grid node power consumption and electricity carbon emission factors, determine the grid node location of the renewable energy project;

[0133] According to the newly added renewable energy installed capacity, the newly added renewable energy power source node is placed at node x, and the regional power carbon emission model is established, which is:

[0134]

[0135] In the formula, Γ area,t is the total electricity carbon emissions of the region in period t, c a,t , Q a,t are the average grid carbon emission factor and load power of node a in period t, x is the node where the newly added renewable energy source is located, and 1≤x≤A, c x,t is the average grid carbon emission factor of node x during period t.

[0136] The newly added renewable energy power sources are placed at Node 1, Node 2, ..., Node A respectively, and the regional electricity carbon emissions under each scenario are calculated. The scenario with the smallest regional electricity carbon emissions is taken as the grid node location where the newly added renewable energy power source is located.

[0137] Embodiment 2:

[0138] See also Figure 2 , a regional renewable energy configuration system based on electricity carbon emissions, comprising:

[0139] Regional source-charge imbalance prediction model building module: used to collect historical power generation and load data of the target area and establish a regional source-charge imbalance prediction model;

[0140] 1.1 The regional source-charge imbalance prediction model constructed is:

[0141] Q balance,t =Q source,t -Q load,t

[0142] In the formula, Q balance,t , Q source,t , Q load,t They are the unbalanced power, generated power and load power in the target area during period t respectively.

[0143] 1.2 Power generation Q source,t The predicted values ​​are:

[0144]

[0145] Where K is the number of regional power types, Q k,t is the predicted value of the power generation of the kth power source in the region during period t.

[0146] The values ​​of the Spearman correlation coefficient are:

[0147]

[0148] Where ρ is the Spearman correlation coefficient, R(x j )、R(y j ) are the positions of variables x and y respectively, are the average ranks of variables x and y respectively, and m is the number of samples.

[0149] The Stacking algorithm is used to calculate the power generation Q of different power types in the target area. k,t To make a prediction, the prediction steps are:

[0150] 1.2.1 Use the Spearman correlation coefficient method to analyze the correlation between power generation influencing factors and power generation of different types of power sources. Rank the power generation influencing factors based on the Spearman correlation coefficient, select the historical data of power generation influencing factors with the top n correlation rankings, and combine them with the historical power generation to form the initial data set of different power types. Perform steps 1.2.2-1.2.4 for each power type;

[0151] 1.2.2 Divide the initial data set into a meta-training set and a meta-test set on the meta-learner, and divide the meta-training set into a base training set and a base test set on the base learner;

[0152] 1.2.3 Train and predict the base learners. Select ResNet, VGG, and MobileNet algorithm models as base learners. Use K-fold cross-validation to train the model on K-1 folds and verify it on the Kth fold. Each base learner is trained K times, for a total of 3K times, to generate all base models and prediction results.

[0153] 1.2.4 Based on the prediction data of the base learner, a prediction model is made on the meta-learner. Linear regression and logistic regression algorithms are used as meta-learners respectively. The training results of the two learners are averaged to obtain the final prediction result.

[0154] 1.3 Regional power load forecasting is carried out through the multi-layer perceptron model based on empirical mode decomposition. The specific steps are:

[0155] 1.3.1 Construct a regional historical load power data set, divide the regional historical load power data set into a training set and a test set, and use the empirical mode decomposition algorithm to decompose the historical load power data into n intrinsic mode components and 1 residual:

[0156]

[0157] In the formula, Q load 、IMF i , Res are the historical load power, the i-th intrinsic mode component, and the residual respectively;

[0158] 1.3.2 The decomposed intrinsic mode components are reconstructed using the extreme point partitioning method to form high-frequency, medium-frequency and low-frequency components. A multi-layer perceptron prediction model is established for each frequency band, namely HF-MLP, IF-MLP and LF-MLP;

[0159] 1.3.3 The Spearman correlation coefficient method is used to analyze the correlation between candidate features and high-frequency, medium-frequency, and low-frequency components, and the components with absolute values ​​of correlation coefficients greater than the set threshold are selected as the inputs of the HF-MLP, IF-MLP, and LF-MLP models;

[0160] 1.3.4 Use the training set to train and adjust the parameters of the HF-MLP, IF-MLP, and LF-MLP models. After the model is trained, the high-frequency, medium-frequency, and low-frequency components of the power load data are predicted on the test set, and the prediction results are superimposed to obtain the load power prediction value Q load,t .

[0161] Renewable energy investment capacity model building module: used to build a renewable energy investment model with the goal of minimizing the total cost of regional power generation and determine the newly added renewable energy installed capacity;

[0162] Considering the cost of new renewable energy power generation P invest,t , the cost of electricity generation from existing renewable energy sources P genera,t 、The cost of electricity transferred from other regions P buy,t And converted into green certificate income P green,t , with the total regional power generation cost P all,t Minimum goal: building a renewable energy investment model:

[0163] minP all,t =min(P invest,t +P genera,t +P buy,t -P green,t )

[0164]

[0165] P buy,t =-Q balance,t ×W t

[0166]

[0167] Where R is the type of new renewable energy source; G r,t , U r,t , H r,t , r,t are the installed capacity, unit installed capacity cost, power generation utilization hours, and unit power operation and maintenance cost coefficient of the rth type of newly added renewable energy power source in period t; b is the discount rate; T r Add the service life of renewable energy source for category r; V k,t W is the electricity cost of the kth type of stock power source in the region during period t; t is the inter-regional electricity transaction unit price during period t; S is the type of renewable energy in the stock power source; Q s,t is the power generation of the sth stock of renewable energy in period t; δ is the quantitative coefficient of renewable energy power generation converted into green certificates, I t It is the average transaction price of green certificates during period t.

[0168] Regional node power carbon emission factor library construction module: used to establish a regional node power carbon emission factor library based on power flow information;

[0169] The node power carbon emission factor is obtained based on the power flow, which is the average carbon emission factor of node a in the target area during period t, and is:

[0170]

[0171] In the formula, c a is the average grid carbon emission factor of regional node a, is the direct carbon emission of the power plant directly connected to node a, is the indirect carbon emission injected into the carbon emission flow of node a, Q a is the load power of node a, Q out is the power flow out of node a; the target area has A nodes, and the power carbon emission factor library has {c1, c2, ..., c A}.

[0172] Renewable energy configuration plan generation module: It is used to determine the grid node location of the renewable energy project with the goal of minimizing regional carbon emissions, combined with the regional grid node power consumption and electricity carbon emission factors.

[0173] According to the newly added renewable energy installed capacity, the newly added renewable energy power source node is placed at node x, and the regional power carbon emission model is established, which is:

[0174]

[0175] In the formula, Γ area,t is the total electricity carbon emissions of the region in period t, c a,t , Q a,t are the average grid carbon emission factor and load power of node a in period t, x is the node where the newly added renewable energy source is located, and 1≤x≤A, c x,t is the average grid carbon emission factor of node x during period t;

[0176] The newly added renewable energy power sources are placed at Node 1, Node 2, ..., Node A respectively, and the regional electricity carbon emissions under each scenario are calculated. The scenario with the smallest regional electricity carbon emissions is taken as the grid node location where the newly added renewable energy power source is located.

Claims

1. A method for configuring regional renewable energy based on electricity carbon emissions, characterized in that: include: S1. Collect historical power generation and load data of the target area, establish a regional source-load imbalance prediction model, and determine the target area source-load imbalance prediction value; S2. Determine the newly added renewable energy installed capacity based on the predicted value of source-load imbalance in the target area; S3. Establish a regional node power carbon emission factor library based on power flow information; S4. Based on the newly added installed capacity of renewable energy, with the goal of minimizing regional carbon emissions, combined with the regional grid node power consumption and electricity carbon emission factors, determine the grid node location of the renewable energy project.

2. A method for configuring regional renewable energy based on electricity carbon emissions according to claim 1, characterized in that: In S1, the regional source-charge imbalance prediction model constructed in 1.1 is: Q balance,t =Q source,t -Q load,t In the formula, Q balance,t , Q source,t , Q load,t They are the unbalanced power, generated power and load power in the target area during period t respectively.

3. A method for configuring regional renewable energy based on electricity carbon emissions according to claim 2, characterized in that: In S1, 1.2 generated electricity Q source,t The predicted values ​​are: Where K is the number of regional power types, Q k,t is the predicted value of the power generation of the kth power source in the region during period t.

4. A method for configuring regional renewable energy based on electricity carbon emissions according to claim 3, characterized in that: In S1, the values ​​of the Spearman correlation coefficient are: Where ρ is the Spearman correlation coefficient, R(x j )、R(y j ) are the positions of variables x and y respectively, are the average ranks of variables x and y respectively, and m is the number of samples.

5. A method for configuring regional renewable energy based on electricity carbon emissions according to claim 3 or 4, characterized in that: In S1, the stacking algorithm is used to calculate the power generation Q of different power types in the target area. k,t To make a prediction, the prediction steps are: 1.2.1 Use the Spearman correlation coefficient method to analyze the correlation between power generation influencing factors and power generation of different types of power sources. Rank the power generation influencing factors based on the Spearman correlation coefficient, select the historical data of power generation influencing factors with the top n correlation rankings, and combine them with the historical power generation to form the initial data set of different power types. Perform steps 1.2.2-1.2.4 for each power type; 1.2.2 Divide the initial data set into a meta-training set and a meta-test set on the meta-learner, and divide the meta-training set into a base training set and a base test set on the base learner; 1.2.3 Train and predict the base learners. Select ResNet, VGG, and MobileNet algorithm models as base learners. Use K-fold cross-validation to train the model on K-1 folds and verify it on the Kth fold. Each base learner is trained K times, for a total of 3K times, to generate all base models and prediction results. 1.2.4 Based on the prediction data of the base learner, a prediction model is made on the meta-learner. Linear regression and logistic regression algorithms are used as meta-learners respectively. The training results of the two learners are averaged to obtain the final prediction result.

6. A method for configuring regional renewable energy based on electricity carbon emissions according to claim 5, characterized in that, in said S1, 1.3 performs regional electricity load forecasting through a multi-layer perceptron model based on empirical mode decomposition, and the specific steps are: 1.3.1 Construct a regional historical load power data set, divide the regional historical load power data set into a training set and a test set, and use the empirical mode decomposition algorithm to decompose the historical load power data into n intrinsic mode components and 1 residual: In the formula, Q load 、IMF i , Res are the historical load power, the i-th intrinsic mode component, and the residual respectively; 1.3.2 The decomposed intrinsic mode components are reconstructed using the extreme point partitioning method to form high-frequency, medium-frequency and low-frequency components. A multi-layer perceptron prediction model is established for each frequency band, namely HF-MLP, IF-MLP and LF-MLP; 1.3.3 The Spearman correlation coefficient method is used to analyze the correlation between candidate features and high-frequency, medium-frequency, and low-frequency components, and the components with absolute values ​​of correlation coefficients greater than the set threshold are selected as the inputs of the HF-MLP, IF-MLP, and LF-MLP models; 1.3.4 Use the training set to train and adjust the parameters of the HF-MLP, IF-MLP, and LF-MLP models. After the model is trained, the high-frequency, medium-frequency, and low-frequency components of the power load data are predicted on the test set, and the prediction results are superimposed to obtain the load power prediction value Q load,t .

7. A method for configuring regional renewable energy based on electricity carbon emissions according to claim 6, characterized in that in said S2: based on the target area source-charge unbalanced power prediction value Q balance,t , considering the cost of new renewable energy power generation P invest,t , the cost of electricity generation from existing renewable energy sources P genera,t 、The cost of electricity transferred from other regions P buy,t And converted into green certificate income P green,t , with the total regional power generation cost P all,t Minimum goal: building a renewable energy investment model: minP all,t =min(P invest,t +P genera,t +P buy,t -P green,t ) P buy,t =-Q balance,t ×W t Where R is the type of new renewable energy source; G r,t , U r,t , H r,t , r,t are the installed capacity, unit installed capacity cost, power generation utilization hours, and unit power operation and maintenance cost coefficient of the rth type of newly added renewable energy power source in period t; b is the discount rate; T r Add the service life of renewable energy source for category r; V k,t W is the electricity cost of the kth type of stock power source in the region during period t; t is the inter-regional electricity transaction unit price during period t; S is the type of renewable energy in the stock power source; Q s,t is the power generation of the sth stock of renewable energy in period t; δ is the quantitative coefficient of renewable energy power generation converted into green certificates, I t It is the average transaction price of green certificates during period t.

8. A method for configuring regional renewable energy based on electricity carbon emissions according to claim 7, characterized in that: In S3, the node power carbon emission factor is obtained based on the power flow, which is the average carbon emission factor of node a in the target area during period t, and is: In the formula, c a is the average grid carbon emission factor of regional node a, is the direct carbon emission of the power plant directly connected to node a, is the indirect carbon emission injected into the carbon emission flow of node a, Q a is the load power of node a, Q out is the power flow out of node a; the target area has A nodes, and the power carbon emission factor library has {c1, c2, ..., c A }.

9. A method for configuring regional renewable energy based on electricity carbon emissions according to claim 8, characterized in that: In S4: According to the newly added renewable energy installed capacity, the newly added renewable energy power source node is placed at node x, and the regional power carbon emission model is established, which is: In the formula, Γ area,t is the total electricity carbon emissions of the region in period t, c a,t , Q a,t are the average grid carbon emission factor and load power of node a in period t, x is the node where the newly added renewable energy source is located, and 1≤x≤A, c x,t is the average grid carbon emission factor of node x during period t; The newly added renewable energy power sources are placed at Node 1, Node 2, ..., Node A respectively, and the regional electricity carbon emissions under each scenario are calculated. The scenario with the smallest regional electricity carbon emissions is taken as the grid node location where the newly added renewable energy power source is located.

10. A regional renewable energy configuration system based on electricity carbon emissions, characterized in that: The renewable energy configuration system is used to execute a regional renewable energy configuration method based on electricity carbon emissions as described in any one of claims 1 to 9: The renewable energy configuration system comprises: Regional source-charge unbalanced prediction model building module: used to collect historical power generation and load data of the target area, establish a regional source-charge unbalanced prediction model, and determine the target area source-charge unbalanced power prediction value; Renewable energy investment model building module: used to determine the newly added renewable energy installed capacity based on the target area source-load imbalance power forecast value; Regional node power carbon emission factor library construction module: used to establish a regional node power carbon emission factor library based on power flow information; Renewable energy configuration plan generation module: used to determine the grid node location of renewable energy projects based on the newly added renewable energy installed capacity, with the goal of minimizing regional carbon emissions, combined with regional grid node power consumption and electricity carbon emission factors.

Images