Multi-level collaborative protection method for offshore wind power direct current sending-out system

Abstract

The invention discloses a multi-level collaborative protection method for an offshore wind power direct current sending-out system, and relates to the technical field of offshore wind power direct current transmission, and the method specifically comprises the steps: 1, collecting the historical safe operation data of the offshore wind power direct current sending-out system at different levels, and 2, carrying out the multi-level collaborative protection of the offshore wind power direct current sending-out system based on the historical safe operation data; the method comprises the following steps: step 1, establishing a safe operation data prediction model constructed based on an LSTM network, and step 2, carrying out protection and early warning on the offshore wind power DC transmission system through the trained prediction model, and according to historical data in a normal working state, carrying out prediction on multi-level data of the offshore wind power DC transmission system through the LSTM network. Whether early warning information is sent or not is judged according to the error range of the predicted value and the true value, and the safety and stability of the offshore wind power direct current sending-out system are remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of offshore wind power direct current transmission, and in particular relates to a multi-level coordinated protection method for an offshore wind power direct current transmission system. Background Art

[0002] With the growth of global energy demand and the enhancement of environmental awareness, offshore wind power has developed rapidly as a clean and renewable energy form. Large-scale offshore wind farms use DC transmission technology to transmit electricity over long distances, which has the advantages of high transmission efficiency and low line loss. However, the offshore wind power DC transmission system faces complex operating environments and various fault challenges, such as line faults, converter faults, DC bus faults, etc. These faults may cause system instability, equipment damage, or even large-scale power outages, posing a serious threat to the safe and stable operation of the power system.

[0003] Most existing protection methods are based on single-level fault detection and isolation, lacking a multi-level, systematic collaborative protection mechanism. In complex and changeable fault situations, these methods often fail to quickly and accurately locate faults and take corresponding protection measures, leading to the expansion of faults and affecting the overall operation of the system. Summary of the invention

[0004] The purpose of the present invention is to provide a multi-level coordinated protection method for an offshore wind power DC transmission system, so as to solve the technical problem in the prior art that an offshore wind power DC transmission system lacks multi-level and systematic coordinated protection.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] The multi-level coordinated protection method for offshore wind power DC transmission system includes:

[0007] Step 1: Collect historical safety operation data of offshore wind power DC transmission systems at different levels.

[0008] By pre-installing sensors and monitoring equipment in offshore wind farms, the time points are determined at intervals of time t, and the monitoring data obtained by the sensors and monitoring equipment at each time point is collected to achieve real-time collection of data at the equipment layer, system layer and power grid layer of offshore wind farms. The equipment layer data mainly includes the output power, voltage and current of important power generation equipment in offshore wind farms. Important power generation equipment includes: wind turbines and permanent magnet synchronous generators;

[0009] System-level data mainly include: input voltage and current on the AC side of the converter, output voltage and current on the DC side of the converter;

[0010] The grid layer data mainly include: grid load, grid frequency and voltage stability;

[0011] Through time series analysis, determine the device layer data, system layer data, and grid layer data within the same time period, and upload them to the data processing cloud platform;

[0012] Record the time period when the offshore wind power DC transmission system is operating normally, no device failure information appears, and the grid is operating stably as the safe time period. Collect the device layer data, system layer data, and grid layer data within N historical safe time periods in the data processing cloud platform, which are collectively referred to as historical safe operation data. Denote the device layer data, system layer data, and grid layer data as different types of historical safe operation data.

[0013] Step 2: Based on the historical safe operation data, establish a safe operation data prediction model based on the LSTM network structure.

[0014] Clean the historical safe operation data, remove invalid, abnormal, or missing data points, sort them according to the historical safe time period to obtain the historical safe operation data set. Perform normalization processing on the historical safe operation data set. Select the historical safe operation data included in O historical safe time periods as the control set, and select the historical safe operation data included in Y historical safe time periods as the training set. Among them, the historical safe time period of the training set occurs before the historical safe time period of the control set, and O + Y < N. Classify and input them into the initialized LSTM network for training. Input the device layer data included in different time periods in the training set into the LSTM network for training and predicting the device layer data, input the system layer data in different time periods into the LSTM network for training and predicting the system layer data, and input the grid layer data in different time periods into the LSTM network for training and predicting the grid layer data;

[0015] The LSTM network structure includes an input control gate, an output control gate, a forget control gate, a memory cell, and a recurrent unit; after the input training set is screened by the input control gate, it is retained in the memory cell. The memory cell is connected to the recurrent unit, enabling the data in the memory cell to achieve a linear feedback loop through the recurrent unit. According to the preset forget control gate weight parameter, the output control gate determines whether the data is output. The output response function of each control gate is jointly determined by the sigmoid function, the input vector, the hidden state vector, and the weight matrix. The weight matrix includes: the bias weight, the input weight for different control gates, and the recurrent weight unique to the input control gate, the forget control gate, and the output control gate. The update mechanism of the internal state of the LSTM model is specifically as follows:

[0016] h k = Ou(x k , h k-1 ) * σ(C k );

[0017] Among them, h k represents the predicted value at the kth time point, Ou represents the output function of the output control layer, σ(·) represents the sigmoid function, C k represents the memory cell update mechanism at the kth time point, x k Represents the input data at the kth time point. For an LSTM network, only one type of data can be input. For example, an LSTM network that predicts device-level data can only input device-level data, an LSTM network that predicts system-level data can only input system-level data, and an LSTM network that predicts power grid-level data can only input power grid-level data.

[0018] Furthermore, for the memory cell update mechanism at the kth time point, the specific formula is as follows:

[0019] C k =fo(x k ,h k )*C k-1 +in(x k ,h k-1 )*σ(·);

[0020] Among them, C k represents the memory cell update mechanism at the kth time point, fo represents the output function of the forget control gate, in represents the output function of the input control gate, C k-1 represents the memory cell updating mechanism at the k-1th time point;

[0021] Different types of historical safety operation data are predicted through different types of LSTM networks, and the predicted values ​​of the prediction model are compared with the true values ​​in the control set. The specific method is as follows:

[0022]

[0023] Among them, R represents the prediction accuracy of the prediction model, M means that the prediction is made at M time points, and the smaller the value of R is, the better the prediction performance of the model is. k represents the predicted value at the kth time point, Represents the true value in the control set at the kth time point, sets the prediction accuracy threshold D, and when the prediction model continuously predicts R of M time points less than or equal to D, the prediction model training is judged to be completed.

[0024] Step 3: Use the trained prediction model to provide protection warning for the offshore wind power DC transmission system.

[0025] Based on the equipment layer data, system layer data and power grid layer data of the offshore wind power DC transmission system at historical time points, the equipment layer data, system layer data and power grid layer data of the offshore wind power DC transmission system at the current time point are predicted, and the error ranges of the predicted values ​​and the actual values ​​of the equipment layer data, system layer data and power grid layer data at the current time point are compared, and the equipment layer error threshold, the system layer error threshold and the power grid layer error threshold are set respectively. If the error between the predicted value and the actual value of different types of historical safety operation data exceeds the corresponding error threshold, the warning information of the historical safety operation data of this type is triggered, and corresponding response measures are taken according to the warning information of different types of historical safety operation data, for example: taking measures to adjust the power generation strategy for the equipment layer warning information, taking measures to check the converter status for the system layer warning information, and taking measures to adjust the power grid dispatch for the power grid layer warning information.

[0026] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0027] 1. The multi-level collaborative protection method of the present invention significantly improves the safety and stability of the offshore wind power DC transmission system. Through real-time monitoring and early prediction, it can quickly locate and restore early warning information, effectively reducing the impact of faults on the system. At the same time, the LSTM network can continuously optimize the prediction accuracy based on historical data and real-time data, and improve the accuracy and reliability of early warning information. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0029] Figure 1 A step diagram of a multi-level coordinated protection method for an offshore wind power DC transmission system is shown; DETAILED DESCRIPTION

[0030] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] like Figure 1 The multi-level coordinated protection method for the offshore wind power DC transmission system shown in the figure specifically includes the following steps:

[0032] Step 1: Collect historical safe operation data of the offshore wind power DC transmission system at different levels.

[0033] Through sensors and monitoring devices pre-installed in the offshore wind farm, determine time points at intervals of time t, and collect the monitoring data obtained by the sensors and monitoring devices at each time point to achieve real-time acquisition of data at the equipment layer, system layer, and grid layer of the offshore wind farm. The equipment layer data mainly includes the output power, voltage, and current of important power generation equipment in the offshore wind farm. The important power generation equipment includes: wind turbines and permanent magnet synchronous generators;

[0034] The system layer data mainly includes: the input voltage and current on the AC side of the converter, and the output voltage and current on the DC side of the converter;

[0035] The grid layer data mainly includes: grid load, grid frequency, and voltage stability;

[0036] Through time series analysis, determine the equipment layer data, system layer data, and grid layer data within the same time period, and upload them to the data processing cloud platform;

[0037] Record the time period when the offshore wind power DC transmission system operates normally, the equipment does not have fault information, and the grid operates stably as a safe time period. Collect the equipment layer data, system layer data, and grid layer data within N historical safe time periods in the data processing cloud platform, which are collectively referred to as historical safe operation data. Denote the equipment layer data, system layer data, and grid layer data as different types of historical safe operation data.

[0038] Step 2: Based on the historical safe operation data, establish a safe operation data prediction model based on the LSTM network structure.

[0039] Clean the historical safe operation data, remove invalid, abnormal, or missing data points, classify and sort them according to the historical safe time periods to obtain a historical safe operation data set. Perform normalization processing on the historical safe operation data set. Select the historical safe operation data included in O historical safe time periods as the control set, and select the historical safe operation data included in Y historical safe time periods as the training set. The historical safe time periods of the training set occur before the historical safe time periods of the control set, and O + Y < N. Classify and input them into the initialized LSTM network for training. Input the equipment layer data included in different time periods in the training set into the LSTM network for training and predicting equipment layer data, use the system layer data in different time periods for the LSTM network for training and predicting system layer data, and input the grid layer data in different time periods into the LSTM network for training and predicting grid layer data;

[0040] The LSTM network structure includes an input control gate, an output control gate, a forget control gate, a memory cell, and a recurrent unit. After the input training set is screened by the input control gate, it is retained in the memory cell. The memory cell is connected to the recurrent unit so that the data in the memory cell can realize a linear feedback loop through the recurrent unit. According to the preset forget control gate weight parameter, the output control gate determines whether the data is output. The output response function of each control gate is jointly determined by the sigmoid function, the input vector, the hidden state vector, and the weight matrix. The weight matrix includes: bias weights and input weights for different control gates, as well as the input control gate, forget control gate, and output control gate-specific recurrent weights. The update mechanism of the internal state of the LSTM model is as follows:

[0041] h k =Ou(x k ,h k-1 )*σ(C k );

[0042] Among them, h k represents the predicted value at the kth time point, Ou represents the output function of the output control layer, σ(·) represents the sigmoid function, C k represents the memory cell update mechanism at the kth time point, x k Represents the input data at the kth time point. For an LSTM network, only one type of data can be input. For example, an LSTM network that predicts device-level data can only input device-level data, an LSTM network that predicts system-level data can only input system-level data, and an LSTM network that predicts power grid-level data can only input power grid-level data.

[0043] Furthermore, for the memory cell update mechanism at the kth time point, the specific formula is as follows:

[0044] C k =fo(x k ,h k )*C k-1 +in(x k ,h k-1 )*σ(·);

[0045] Among them, C k represents the memory cell update mechanism at the kth time point, fo represents the output function of the forget control gate, in represents the output function of the input control gate, C k-1 represents the memory cell updating mechanism at the k-1th time point;

[0046] Different types of historical safety operation data are predicted through different types of LSTM networks, and the predicted values ​​of the prediction model are compared with the true values ​​in the control set. The specific method is as follows:

[0047]

[0048] Among them, R represents the prediction accuracy of the prediction model, M means that the prediction is made at M time points, and the smaller the value of R is, the better the prediction performance of the model is. k represents the predicted value at the kth time point, Represents the true value in the control set at the kth time point, sets the prediction accuracy threshold D, and when the prediction model continuously predicts R of M time points less than or equal to D, the prediction model training is judged to be completed.

[0049] Step 3: Use the trained prediction model to provide protection warning for the offshore wind power DC transmission system.

[0050] Based on the equipment layer data, system layer data and power grid layer data of the offshore wind power DC transmission system at historical time points, the equipment layer data, system layer data and power grid layer data of the offshore wind power DC transmission system at the current time point are predicted, and the error ranges of the predicted values ​​and the actual values ​​of the equipment layer data, system layer data and power grid layer data at the current time point are compared, and the equipment layer error threshold, the system layer error threshold and the power grid layer error threshold are set respectively. If the error between the predicted value and the actual value of different types of historical safety operation data exceeds the corresponding error threshold, the warning information of the historical safety operation data of this type is triggered, and corresponding response measures are taken according to the warning information of different types of historical safety operation data, for example: taking measures to adjust the power generation strategy for the equipment layer warning information, taking measures to check the converter status for the system layer warning information, and taking measures to adjust the power grid dispatch for the power grid layer warning information.

[0051] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

[0052] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A multi-level coordinated protection method for an offshore wind power DC transmission system, characterized in that: Including: Step 1: Collect historical safe operation data of the offshore wind power DC transmission system at different levels; Record the time period when the offshore wind power DC transmission system operates normally, no equipment failure information appears, and the power grid operates stably as the safe time period. The historical safe operation data is the data within the safe time period; Step 2: Based on the historical safe operation data, establish a safe operation data prediction model constructed based on the LSTM network; Through different types of historical safe operation data, train an LSTM network that can predict different types of historical safe operation data; Step 3: Through the trained prediction model, conduct protection early warning on the offshore wind power DC transmission system.

2. The multi-level coordinated protection method for an offshore wind power DC transmission system according to claim 1 is characterized in that: Collect historical safe operation data of the offshore wind power DC transmission system at different levels. The specific method is: Through sensors and monitoring devices pre-installed in the offshore wind farm, determine time points at intervals of time t, and collect the monitoring data obtained by the sensors and monitoring devices at each time point to realize the real-time acquisition of data at the equipment layer, system layer, and power grid layer of the offshore wind farm; The equipment layer data mainly includes the output power, voltage, and current of important power generation equipment in the offshore wind farm; The system layer data mainly includes: the input voltage and current on the AC side of the converter, and the output voltage and current on the DC side of the converter; The power grid layer data mainly includes: power grid load, power grid frequency, and voltage stability; Collect the equipment layer data, system layer data, and power grid layer data within N historical safe time periods in the data processing cloud platform, and collectively call them historical safe operation data.

3. The multi-level coordinated protection method for an offshore wind power DC transmission system according to claim 1 is characterized in that: Establish a safe operation data prediction model constructed based on the LSTM network. The specific method is: Select the historical safe operation data included in O historical safe time periods as the control set, and select the historical safe operation data included in Y historical safe time periods as the training set. Among them, the historical safe time periods of the training set occur before the historical safe time periods of the control set, and O + Y < N, and classify and input them into the initialized LSTM network for training; Establish an LSTM network structure including input control gate, output control gate, forget control gate, memory cell and recurrent unit, and use h k =Ou(x k ,h k-1 )*σ(C k ) determines the update mechanism of the internal state of the LSTM model, where h k represents the predicted value at the kth time point, Ou represents the output function of the output control layer, σ(·) represents the sigmoid function, C k represents the memory cell update mechanism at the kth time point, x k Represents the input data at the kth time point, thereby establishing a safe operation data prediction model based on the LSTM network. Through multiple predictions, the error between the predicted value and the true value is compared to determine whether the prediction model is trained.

4. The multi-level coordinated protection method for an offshore wind power DC transmission system according to claim 1 is characterized in that: The memory cell update mechanism at the kth time point specifically includes: Using formula C k =fo(x k ,h k )*C k-1 +in(x k ,h k-1 )*σ(·) represents the memory cell updating mechanism at the kth time point, where C k represents the memory cell update mechanism at the kth time point, fo represents the output function of the forget control gate, in represents the output function of the input control gate, C k-1 Represents the memory cell updating mechanism at the k-1th time point.

5. The multi-level coordinated protection method for an offshore wind power DC transmission system according to claim 1 is characterized in that: Through multiple predictions, compare the error between the predicted value and the true value to judge whether the prediction model is trained. The specific method is: By formula The predicted value of the prediction model is compared with the true value in the control set, where R represents the prediction accuracy of the prediction model, M represents the prediction of M time points, and the smaller the value of R is, the better the prediction performance of the model is. k represents the predicted value at the kth time point, h~ k Represents the true value in the control set at the kth time point, sets the prediction accuracy threshold D, and when the prediction model continuously predicts R of M time points less than or equal to D, the prediction model training is judged to be completed.

6. The multi-level coordinated protection method for an offshore wind power DC transmission system according to claim 1 is characterized in that: Through the trained prediction model, conduct protection early warning on the offshore wind power DC transmission system. The specific method is: Based on the equipment layer data, system layer data, and power grid layer data of the historical time points of the offshore wind power DC transmission system, predict the equipment layer data, system layer data, and power grid layer data of the current time point of the offshore wind power DC transmission system, compare the error ranges between the predicted values and the true values of the equipment layer data, system layer data, and power grid layer data at the current time point, and respectively set the equipment layer error threshold, system layer error threshold, and power grid layer error threshold. If the error between the predicted value and the true value of different types of historical safe operation data exceeds the corresponding error threshold, then trigger the warning information belonging to this type of historical safe operation data.

Images