Voltage prediction method for ac-dc hybrid distribution network considering vsc control strategy

Abstract

The present application relates to a kind of AC-DC hybrid distribution network voltage prediction method considering VSC control strategy, belong to distribution network technical field.The method comprises the following steps: S1: obtaining AC-DC hybrid distribution network operating data including node voltage, node active power, node reactive power, time characteristics, weather characteristics, VSC control strategy and other data, and processing VSC node;S2: correlation analysis is carried out to distribution network operating data and the voltage to be predicted node, selects the data with strong correlation with the voltage to be predicted node as the input sample data of prediction model, and sample data is divided into training set and test set according to proportion;S3: the voltage prediction model considering VSC control strategy is established, and training set data is used for training in the AC side and DC side of hybrid distribution network respectively, and test set data is used to verify the prediction effect.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power distribution networks, and relates to a voltage prediction method for an AC-DC hybrid power distribution network considering a VSC control strategy. BACKGROUND

[0002] In recent years, direct-current power sources and loads such as photovoltaic power and electric vehicle charging piles gradually increase in the proportion in power distribution networks. A direct-current power distribution network is connected to an alternating-current power distribution network through a voltage source converter (VSC) to form an AC-DC hybrid power distribution network, which becomes a future development trend. However, the access of the direct-current power distribution network easily causes changes in power flow distribution of the power distribution network, thereby causing problems such as voltage high-voltage out-of-limit and low-voltage out-of-limit, and affecting voltage stability. Therefore, by predicting node voltages of the power distribution network, voltage out-of-limit situations that are about to occur can be identified in advance, so that the voltage of the power distribution network can be managed in a timely manner. At present, voltage prediction methods for power distribution networks include a BP neural network and an LSTM method. However, the BP neural network lacks modeling capability for internal dependency relationships of data when processing time-series data, and the LSTM can process long-term dependency relationships in data but is relatively weak in feature extraction.

[0003] Therefore, the application provides a voltage prediction model based on the combination of a CNN and an LSTM, which can capture spatial and temporal features of data at the same time, thereby improving the accuracy of the voltage prediction model. The VSC has strong voltage regulation capability and can stabilize or change voltage levels in the power grid by adjusting output voltages, reactive power and active power. Since the control strategy of the VSC directly affects its output behavior, different control strategies will result in different voltage regulation effects. Therefore, if the control strategy of the VSC is not considered when performing voltage prediction, the prediction result may not be consistent with the actual situation.

[0004] In view of this problem, the application provides a voltage prediction method for an AC-DC hybrid power distribution network considering a VSC control strategy. SUMMARY

[0005] Therefore, the application provides a voltage prediction method for an AC-DC hybrid power distribution network considering a VSC control strategy, which first acquires power flow operation data of the AC-DC hybrid power distribution network, then establishes neural network models on the alternating-current side and the direct-current side according to the operation data of the power distribution network, and predicts node voltages.

[0006] To achieve the above-mentioned purpose, the application provides the following technical scheme.

[0007] The voltage prediction method for the AC-DC hybrid power distribution network considering the VSC control strategy comprises the following steps.

[0008] S1: Obtain AC / DC hybrid distribution network operation data including node voltage, node active power, node reactive power, time characteristics, weather characteristics, and data of a voltage source converter (VSC) control strategy, and process the VSC node;

[0009] S2: Perform correlation analysis on the distribution network operation data and the node voltage to be predicted, select data having strong correlation with the node voltage to be predicted as prediction model input sample data, and divide the sample data into a training set and a test set according to a proportion;

[0010] S3: Establish a voltage prediction model considering the VSC control strategy, and train the model using the training set data on the AC side and the DC side of the hybrid distribution network, respectively, and verify the prediction effect using the test set data.

[0011] Further, in S1, the weather characteristics include temperature, humidity, human comfort, and weather phenomena.

[0012] Further, in S1, the control strategy of the VSC is represented by a number as follows:

[0013] ① represents fixed DC voltage and fixed AC reactive power;

[0014] ② represents fixed DC voltage and fixed AC voltage;

[0015] ③ represents fixed AC active power and fixed AC reactive power;

[0016] ④ represents fixed AC active power and fixed AC voltage;

[0017] ⑤ represents V-P droop control and fixed AC reactive power;

[0018] ⑥ represents V-P droop control and fixed AC voltage;

[0019] ⑦ represents V-I droop control and fixed AC reactive power;

[0020] ⑧ represents V-I droop control and fixed AC voltage.

[0021] Further, in S1, the VSC node is processed, including:

[0022] Predict the node voltage on the AC side:

[0023] When the VSC control strategy is ①, ③, ⑤, or ⑦, the VSC node voltage is used as the model input feature;

[0024] When the VSC control strategy is ②, ④, ⑥, or ⑧, the VSC node reactive power is used as the model input feature;

[0025] Predicting the DC side node voltage:

[0026] When the control strategy is ①, ②, ⑤, ⑦, ⑥ or ⑧, the VSC node active power is taken as the model input feature;

[0027] When the control strategy is ③ or ④, the VSC node voltage is taken as the model input feature.

[0028] Further, in S2, the power distribution network operation data and the node voltage are analyzed for correlation, specifically including:

[0029] (1) Determine the reference sequence

[0030] The reference sequence is fixed sequence data, i.e. the voltage data of the node to be tested, which is used for correlation calculation with other data. The reference sequence X0 is represented as:

[0031] X0 = (x0(1), x0(2), x0(3), …, x0(n))

[0032] In the formula, n represents the number of samples of the sequence, and 0 represents the fixed sequence.

[0033] (2) Determine the comparison sequence

[0034] When there are m types of comparison sequences, each type has n samples, then the comparison sequence X i is represented as:

[0035]

[0036] (3) Data normalization processing

[0037] The original data is normalized by using the mean method, and the formula is:

[0038]

[0039] In the formula, k is the kth sample data in type i;

[0040] (4) Calculate the grey correlation coefficient

[0041] The normalized data is calculated for the grey correlation coefficient:

[0042]

[0043] In the formula, k is the kth sample data in type i, and δ represents the resolution coefficient, which takes a value in the interval (0, 1). The smaller the value, the higher the recognition degree, and it is taken as 0.5;

[0044] (5) Calculate the grey correlation degree

[0045] The grey correlation degree r iThe overall correlation degree of the reference sequence and the comparison sequence is represented as:

[0046]

[0047] (6) The absolute value of the grey correlation degree is sorted from large to small:

[0048] The absolute value of the grey correlation degree is sorted from large to small:

[0049] r rank = rank(|r1|, |r2|, …, |r m |)

[0050] In the formula, |r m | represents the absolute value of the grey correlation degree of the to-be-tested voltage and other data, rank() represents sorting the absolute value of the grey correlation degree from large to small, and r rank represents the sorting result.

[0051] Further, in the S3, a voltage prediction model considering the VSC control strategy is established, and specifically includes:

[0052] S31: Obtain operation data of the AC-DC hybrid distribution network, including node voltage data, node active power data, node reactive power data, VSC control strategy, and weather feature data;

[0053] S32: Preprocess the obtained data, including removing outliers, filling missing values, and data normalization;

[0054] S33: Calculate the grey correlation degree of the preprocessed to-be-tested node voltage data and other node data within a certain distance of the node, and sort the absolute values of the grey correlation degrees from large to small; wherein the other node data includes node voltage, node active power, and node reactive power; and sort the absolute values of the grey correlation degrees from large to small;

[0055] S34: According to the sorting result of the absolute values of the grey correlation degrees, select the first k variables with the largest absolute values as the input features of the prediction model, and the value of k is defined by the user; and the to-be-tested node voltage is used as the output of the model;

[0056] S35: Divide the preprocessed data into training set and test set in proportion, and then construct a voltage prediction model based on the combination of CNN and LSTM, and train the model based on the training set on the AC side and the DC side of the hybrid distribution network, and verify the effect of the model using the test set.

[0057] Further, in the S35, the voltage prediction model based on the combination of CNN and LSTM is constructed, wherein the model structure includes 2 convolution layers, 2 pooling layers, 1 LSTM layer, and an activation layer;

[0058] The activation layer adopts a ReLu function, and is expressed as:

[0059] f(a) = max(a, 0)

[0060] In the formula, a represents the input of the activation function, f(a) represents the output of the activation function, and max(a, 0) represents that the activation function takes the maximum value between the input a and 0 as the output.

[0061] Further, in the S35, the model effect is verified, including:

[0062] (1) calculating the mean absolute error (MAE) of the voltage prediction model

[0063]

[0064] In the formula, n is the total number of samples, y is the true value, and y is the model prediction value. i

[0065] (2) calculating the root mean square error (RMSE) of the prediction model

[0066]

[0067] In the formula, n is the total number of samples, y is the true value, and y is the model prediction value. i

[0068] The beneficial effects of the present application are:

[0069] (1) The present application proposes a voltage prediction model based on CNN+LSTM for predicting the voltage of the distribution network. Compared with the traditional BP and LSTM neural network, the voltage prediction model can extract the spatial and temporal characteristics of the data at the same time, thereby improving the prediction accuracy of the voltage prediction model.

[0070] (2) The present application processes the VSC node, selects the to-be-tested node data (including active power and reactive power), other nodes close to the to-be-predicted node, and VSC node data (including voltage, active power and reactive power), weather feature data and time feature data as input variables of the voltage prediction model, and predicts the voltage, thereby enriching the model input features and improving the generalization of the voltage prediction model.

[0071] ​​​​Additional advantages, objects, and features of the application will be apparent to those skilled in the art upon examination of the following specification. It is intended that the application not be limited by any of the details of the specification, unless so expressly desired, but instead be controlled by the full scope of the appended claims. BRIEF DESCRIPTION OF DRAWINGS

[0072] In order to make the objectives, technical solutions and advantages of the present application clearer, the preferred embodiments of the present application will be described in detail below with reference to the drawings, in which:

[0073] Figure 1 is a flow chart of the present application;

[0074] Figure 2 is a schematic diagram of the AC-DC hybrid power distribution network topology;

[0075] Figure 3 is the AC-DC hybrid power distribution network node power flow operation data;

[0076] Figure 4 is a voltage prediction model flow;

[0077] Figure 5 is an AC node voltage prediction result graph;

[0078] Figure 6 is a DC node voltage prediction result graph. DETAILED DESCRIPTION

[0079] The embodiments of the present application will be described in detail below with reference to specific examples. Those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in the specification. The present application can also be implemented or applied in different specific embodiments, and the details in the specification can be modified or changed based on different views and applications without departing from the spirit of the present application. It should be noted that the diagrams provided in the following examples only illustrate the basic concept of the present application in a schematic manner, and the following examples and features in the examples can be combined with each other without conflict.

[0080] The drawings are only used for illustrative explanation, and the representation is only a schematic diagram, not a physical diagram, and should not be understood as a limitation of the present application. In order to better illustrate the embodiments of the present application, some components in the drawings may be omitted, enlarged or reduced, and do not represent the actual size of the product. It is understandable to those skilled in the art that some well-known structures and their descriptions in the drawings may be omitted.

[0081] The same or similar reference numerals in the drawings of the embodiments of the present application correspond to the same or similar components; in the description of the present application, it should be understood that if the orientations or positional relationships indicated by the terms "upper", "lower", "left", "right", "front", "back", etc. are based on the orientations or positional relationships shown in the drawings, they are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, therefore the terms describing the positional relationship in the drawings are only used for exemplary illustration, and cannot be understood as a limitation on the present application, for those skilled in the art, the specific meanings of the above terms can be understood according to the specific circumstances.

[0082] As shown in Figure 1 , the voltage prediction method of the AC-DC hybrid distribution network considering the VSC control strategy includes the following steps:

[0083] S1: Obtain the operation data of the AC-DC hybrid distribution network, including node voltage, node active power, node reactive power, time characteristics, weather characteristics, VSC control strategy, etc., and process the VSC node.

[0084] It should be noted that in the S1, the weather characteristics include temperature, humidity, human comfort, weather phenomena, etc.

[0085] It should be noted that in the S1, the control strategy of the VSC is represented by a number: ① fixed DC voltage and fixed AC reactive power, ② fixed DC voltage and fixed AC voltage, ③ fixed AC active power and fixed AC reactive power, ④ fixed AC active power and fixed AC voltage, ⑤ V-P droop control and fixed AC reactive power, ⑥ V-P droop control and fixed AC voltage, ⑦ V-I droop control and fixed AC reactive power, ⑧ V-I droop control and fixed AC voltage.

[0086] It should be noted that in the S1, the VSC node is processed, including:

[0087] The AC side node voltage is predicted, when the VSC control strategy is ①, ③, ⑤, ⑦, the VSC node voltage is taken as the model input feature, when the VSC control strategy is ②, ④, ⑥, ⑧, the VSC node reactive power is taken as the model input feature; the DC side node voltage is predicted, when the control strategy is ①, ②, ⑤, ⑦, ⑥, ⑧, the VSC node active power is taken as the model input feature, when the control strategy is ③, ④, the VSC node voltage is taken as the model input feature.

[0088] The present embodiment selects the measured operation data of a certain distribution network for analysis, the sampling time of the data is from June 1, 2021 to June 30, 2022, the sampling interval of the data is 1 hour, there are 9480 data, and the schematic diagram of the topology structure of the AC-DC hybrid distribution network is as shown inFigure 2 As shown in the figure, the AC power distribution network is connected with the DC power distribution network through a VSC, and when the VSC control strategy is changed, the AC-DC hybrid power distribution network node flow operation part data such as Figure 3 As shown in the figure.

[0089] S2: Correlation analysis is performed on the power distribution network operation data and the node voltage to be predicted, data having strong correlation with the node voltage to be predicted is selected as the prediction model input sample data, and the sample data is divided into a training set and a test set in proportion.

[0090] It should be noted that the correlation analysis is performed on the power distribution network operation data and the node voltage, and the specific steps include:

[0091] (1) Determine the reference sequence

[0092] The reference sequence is fixed sequence data, that is, the voltage data of the node to be measured, which is used for correlation calculation with other data, and the reference sequence X0 is represented as:

[0093] X0=(x0(1),x0(2),x0(3),…,x0(n))

[0094] In the formula, n represents the sample number of the sequence, and 0 represents the fixed sequence.

[0095] (2) Determine the comparison sequence

[0096] When there are m types of comparison sequences, each type has n samples, and the comparison sequence X i As shown in the formula:

[0097]

[0098] (3) Data normalization processing

[0099] The mean method is used to normalize the original data, and the formula is:

[0100]

[0101] In the formula, k is the kth sample data in type i.

[0102] (4) Calculate the grey correlation coefficient

[0103] The grey correlation coefficient of the normalized data is calculated:

[0104]

[0105] In the formula, k is the kth sample data in type i, and δ represents the resolution coefficient, which is in the value range (0, 1), and the smaller the value, the higher the recognition degree, and the value is 0.5.

[0106] (5) Calculate the gray correlation degree

[0107] Gray correlation degree r i The overall correlation degree of the reference sequence and the comparison sequence is represented as:

[0108]

[0109] (6) Sort the absolute values of the gray correlation degrees from large to small:

[0110] Sort the absolute values of the gray correlation degrees from large to small:

[0111] r rank = rank(|r1|, |r2|, …, |r m |)

[0112] In the formula, |r m | represents the absolute values of the gray correlation degrees of the to-be-tested node voltage and other data, rank() represents sorting the absolute values of the gray correlation degrees from large to small, and r rank represents the sorting result.

[0113] In this example, the gray correlation degrees of the voltage data of other nodes and the to-be-predicted node are calculated, respectively, taking a to-be-predicted node on the AC side and the DC side of the hybrid distribution network as an example. The sorting results of the absolute values of the gray correlation degrees are shown in Table 1.

[0114] Table 1

[0115] AC node variable Grey correlation degree DC node variable Grey correlation degree Adjacent node 1 voltage 0.940271 Adjacent node 2 voltage 0.996213 Adjacent node 3 voltage 0.938829 Adjacent node 1 voltage 0.996154 Adjacent node 2 voltage 0.936875 Adjacent node 3 voltage 0.995913 Temperature and humidity index 0.93274 Adjacent node 1 active power 0.991387 Adjacent node 4 voltage 0.899329 Adjacent node 4 voltage 0.990397 To-be-measured node active power 0.894652 Adjacent node 5 voltage 0.980522 Adjacent node 3 active power 0.864952 To-be-measured node active power 0.979963 … … … …

[0116] S3: Establish a voltage prediction model considering the VSC control strategy, and train the model using the training set data on the AC side and the DC side of the hybrid distribution network, respectively. Meanwhile, verify the prediction effect using the test set data.

[0117] It should be noted that in the S3, the step of establishing a voltage prediction model considering the VSC control strategy includes:

[0118] S31: Obtain the operation data of the AC-DC hybrid distribution network, including node voltage data, node active power data, node reactive power data, VSC control strategy, and weather characteristics data, etc.

[0119] S32: Preprocess the obtained data, including removing outliers, filling missing values, and data normalization.

[0120] S33: Calculate the gray correlation degrees of the preprocessed to-be-tested node voltage data and other node data (node voltage, node active power, and node reactive power) close to the node, and sort the absolute values of the gray correlation degrees from large to small.

[0121] S34: According to the absolute value of the gray correlation degree ranking result, the first k variables with the largest absolute value are selected as the input features of the prediction model, and the value of k is defined by the user, and the voltage of the to-be-tested node is taken as the output of the model;

[0122] S35: The preprocessed data is divided into training set and test set according to the proportion, and a voltage prediction model based on the combination of CNN and LSTM is constructed, and the training set is trained based on the alternating current side and the direct current side of the hybrid distribution network, and the test set is used to verify the effect of the model.

[0123] It should be noted that in the S35, the voltage prediction model based on the combination of CNN and LSTM is constructed, wherein the model structure includes 2 convolutional layers, 2 pooling layers, 1 LSTM layer and an activation layer.

[0124] The activation layer adopts ReLu function, which is represented as:

[0125] f(a) = max(a, 0)

[0126] In the formula, a represents the input of the activation function, f(a) represents the output of the activation function, and max(a, 0) represents the maximum value between the input a and 0 as the output of the activation function.

[0127] It should be noted that in the S35, the effect of the model is verified, including:

[0128] (1) Calculate the mean absolute error (Mean Absolute Error, MAE) of the voltage prediction model

[0129]

[0130] In the formula, n is the total number of samples, y i is the true value, is the predicted value of the model.

[0131] (2) Calculate the root mean square error (Root Mean Square Error, RMSE) of the prediction model

[0132]

[0133] In the formula, n is the total number of samples, y i is the true value, is the predicted value of the model.

[0134] The embodiment takes a to-be-predicted node on the AC side and the DC side of a hybrid power distribution network as an example to perform voltage prediction. First, a neural network prediction model based on CNN+LSTM is constructed, and then the AC side data and the DC side data are used to respectively predict the voltage of the to-be-predicted node. The network structure diagram is as shown in Figure 4 The voltage prediction results of the to-be-predicted nodes on the AC side and the DC side are as shown in Figure 5 , Figure 6 respectively. The prediction model evaluation indexes are as shown in Table 2.

[0135] Table 2

[0136] Data set MAE RMSE AC side training set 0.89 1.03 AC side test set 0.91 1.25 DC side training set 0.87 1.12 DC side test set 0.89 1.23

[0137] Finally, it should be pointed out that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced equivalently without departing from the purpose and scope of the technical solutions, and they should be covered in the scope of the claims of the present application.

Claims

1. A voltage prediction method for AC / DC hybrid distribution networks considering VSC control strategy, characterized in that: The method includes the following steps: S1: Acquire AC / DC hybrid distribution network operation data, including node voltage, node active power, node reactive power, time characteristics, weather characteristics, and data on voltage source converter (VSC) control strategies, and process the VSC nodes; In S1, the VSC control strategy is represented by the following number: ① indicates constant DC voltage and constant AC reactive power; ② Represents constant DC voltage and constant AC voltage; ③ Represents constant AC active power and constant AC reactive power; ④ indicates constant AC active power and constant AC voltage; ⑤ indicates VP droop control and constant AC reactive power; ⑥ indicates VP droop control and constant AC voltage; ⑦ indicates VI droop control and constant AC reactive power; ⑧ indicates VI droop control and constant AC voltage; The VSC node is processed, including: Predicting AC-side node voltages: When the VSC control strategy is ①, ③, ⑤ or ⑦, the VSC node voltage is used as the model input feature; When the VSC control strategy is ②, ④, ⑥ or ⑧, the reactive power of the VSC node is used as the model input feature; Predicting DC-side node voltages: When the control strategy is ①, ②, ⑤, ⑦, ⑥ or ⑧, the active power of the VSC node is used as the model input feature; When the control strategy is ③ or ④, the VSC node voltage is used as the model input feature; S2: Perform correlation analysis between distribution network operation data and the voltage of the node to be predicted, select data that are strongly correlated with the voltage of the node to be predicted as input sample data for the prediction model, and divide the sample data into training set and test set according to the proportion. S3: Establish a voltage prediction model that considers the VSC control strategy, and train it on the AC and DC sides of the hybrid distribution network using training set data respectively, while using test set data to verify the prediction effect.

2. The AC / DC hybrid distribution network voltage prediction method considering VSC control strategy according to claim 1, characterized in that: In S1, weather characteristics include temperature, humidity, human comfort level, and weather phenomena.

3. The AC / DC hybrid distribution network voltage prediction method considering VSC control strategy according to claim 1, characterized in that: In step S2, a correlation analysis is performed between the distribution network operation data and node voltages, specifically including: (1) Determine the reference sequence The reference sequence is fixed-order data, i.e., the voltage data of the node under test, used for correlation calculation with other data. The reference sequence X0 is represented as: X0=(x0(1),x0(2),x0(3),…,x0(n)) In the formula, n represents the number of samples in the sequence, and 0 represents a fixed sequence; (2) Determine the comparison sequence When there are m comparison sequences of different types, where each type contains n samples, then the comparison sequence X... i For example, it can be represented as: (3) Data normalization processing The original data is normalized using the mean method, and the formula is as follows: In the formula, k is the kth sample data in type i; (4) Calculate the grey relational coefficient Calculate the grey relational coefficient for the normalized data: In the formula, k is the kth sample data in type i, δ represents the resolution coefficient, which takes the value range of (0, 1). The smaller the value, the higher the recognition. It is taken as 0.

5. (5) Calculate the grey relational degree Grey relational degree r i This indicates the overall correlation between the reference sequence and the comparison sequence: (6) Ranking by absolute value of grey relational degree: Sort the absolute values ​​of the grey relational degrees from largest to smallest: r rank =rank(|r1|,|r2|,…,|r m |) In the formula, |r m | represents the absolute value of the gray correlation degree between the voltage of the measured transformer area and other data, and rank() represents sorting the absolute values ​​of the gray correlation degree from largest to smallest. rank This indicates the sorting result.

4. The AC / DC hybrid distribution network voltage prediction method considering VSC control strategy according to claim 3, characterized in that: In step S3, a voltage prediction model considering the VSC control strategy is established, specifically including: S31: Acquire AC / DC hybrid distribution network operation data, including node voltage data, node active power data, node reactive power data, VSC control strategy, and weather characteristic data; S32: Preprocess the acquired data, including outlier removal, missing value imputation, and data normalization; S33: Calculate the grey relational degree between the preprocessed voltage data of the node to be measured and the data of other nodes within a certain distance of the node, and sort the absolute values ​​of the grey relational degree from largest to smallest; wherein, the data of other nodes include node voltage, node active power and node reactive power; and sort the absolute values ​​of the grey relational degree from largest to smallest. S34: Based on the sorting results of the absolute values ​​of grey relational degrees, select the k variables with the largest absolute values ​​as the input features of the prediction model. The value of k is defined by the user, and the voltage of the node to be measured is used as the output of the model. S35: Divide the preprocessed data into training and test sets according to the proportion, then construct a voltage prediction model based on the combination of CNN and LSTM, train it on the AC and DC sides of the hybrid distribution network based on the training set, and use the test set to verify the model's performance.

5. The AC / DC hybrid distribution network voltage prediction method considering VSC control strategy according to claim 4, characterized in that: In S35, a voltage prediction model based on the combination of CNN and LSTM is constructed. The model structure includes two convolutional layers, two pooling layers, one LSTM layer, and an activation layer. The activation layer uses the ReLU function, represented as: f(a) = max(a, 0) In the formula, a represents the input of the activation function, f(a) represents the output of the activation function, and max(a,0) means that the activation function takes the maximum value between the input a and 0 as the output.

6. The AC / DC hybrid distribution network voltage prediction method considering VSC control strategy according to claim 5, characterized in that: In step S35, the model's effectiveness is verified, including: (1) Calculate the mean absolute error (MAE) of the voltage prediction model: In the formula, n is the total number of samples, and y i For the true value, These are the model's predicted values; (2) Calculate the root mean square error (RMSE) of the prediction model: In the formula, n is the total number of samples, and y i For the true value, These are the model's predicted values.

Images