A fault identification and early warning protection method for offshore wind farm transmission system based on data-driven state estimation

Abstract

The present invention relates to a method for fault identification and early warning protection of an offshore wind farm transmission system based on data-driven state estimation, comprising: constructing a global state unified model adapted to the heterogeneous topology of offshore wind power, establishing a state estimation model for long-distance transmission cables; collecting cable end measurement data, combining it with the state estimation model of the long-distance transmission cable, and obtaining a state estimate; calculating the residual and fitting error based on the state estimate; judging whether a fault has occurred and performing fault location classification based on the residual and fitting error results, and outputting corresponding protection actions and alarm information. Compared with the prior art, the present invention adopts a three-layer nested architecture of distributed parameter modeling-state solution-residual discrimination, constructs a mapping relationship between a physical model and a state, uses least squares to estimate the state, and integrates the residual amplitude, phase angle change, and spatial distribution characteristics. It can perform high-precision fault identification and location for multiple sections of heterogeneous transmission cables, and provide timely early warning protection.

Description

Technical Field

[0001] The present invention relates to the technical field of offshore wind power, and in particular to a method for fault identification and early warning protection of an offshore wind farm transmission system based on data-driven state estimation. Background Art

[0002] In offshore wind power transmission systems, long-distance high-voltage cables serve as critical transmission pathways for wind farm grid connection. Their operational stability and protection performance are crucial for ensuring the continuity of power transmission and the security of grid regulation. Especially in the context of offshore access and large-capacity grid connection, transmission cable systems generally exhibit strong structural heterogeneity, complex electrical parameters, and highly dynamic operating states. These systems consist of multiple heterogeneous cable sections, including land cables, mudflat sections, and submarine cables. The system topology exhibits high coupling and asymmetry. Furthermore, most wind turbines utilize full-power converters for access, which weakens fault response characteristics. Conventional protection mechanisms struggle to meet the requirements for rapid, accurate, and stable fault identification.

[0003] Current projects still primarily rely on traditional relay protection methods, such as current differential protection and distance protection. These methods rely on measurement indicators such as fault current amplitude and impedance characteristics to establish action criteria. In cable transmission scenarios, the sensitivity and selectivity of traditional protection systems are significantly reduced due to factors such as insufficient support capacity of converter-type wind power sources, sudden changes in inter-section parameters, and unbalanced charging currents. Differential protection relies on communication link synchronization, and communication loss in offshore environments often leads to protection failure or malfunction. Distance protection, on the other hand, is highly sensitive to model parameters and has difficulty handling issues such as impedance distortion and distance measurement deviation in heterogeneous structures, resulting in insufficient overall protection adaptability.

[0004] In recent years, model-driven state estimation protection (MBP) has emerged as a new direction in relay protection research. Based on a physical model of the transmission system and combined with online voltage and current measurements, this approach uses a state estimation algorithm to calculate the system's internal state variables and identifies system anomalies based on the measured and estimated residuals. This approach, which does not rely on current amplitude or traditional impedance models, boasts communication independence, robustness to measurement errors, and structural adaptability. In theory, it is more suitable for the protection of transmission cable systems with complex structures and weak current characteristics. However, most current research remains at the simulation verification level, lacking a universal method adaptable to engineering scenarios. On the one hand, the topological modeling of multiple heterogeneous cables is difficult, limiting the accuracy and convergence of state estimation. On the other hand, the residual judgment criteria remain crude, lacking the ability to systematically identify fault types and disturbance conditions, resulting in insufficient fault location accuracy.

[0005] In summary, traditional differential and distance protection methods rely on the amplitude and direction characteristics of the fault current, making them difficult to adapt to the special operating characteristics of the inverter-type power supply in offshore wind power transmission systems, which have weak current support capabilities and nonlinear short-circuit response.

[0006] The multi-section heterogeneous cable connection structure makes the traditional distance measurement method based on the equivalent impedance model prone to large errors. Especially when there are sudden changes in parameters between different sections, traditional protection methods have difficulty in achieving accurate fault location and phase identification.

[0007] Existing centralized protection systems are highly dependent on communication links and synchronous sampling mechanisms. In actual offshore environments, they face risks such as channel interruption, delay, and reduced accuracy. This can lead to false or non-fault identification, making it difficult to ensure system safety.

[0008] Some state estimation or data-driven methods have not fully integrated the physical structure characteristics of the cable, lack multi-dimensional analysis of model residuals and redundant measurement fault tolerance, and are unable to meet the practical requirements of high-reliability relay protection. Summary of the Invention

[0009] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide a method for fault identification and early warning protection of offshore wind farm transmission systems based on data-driven state estimation, which can perform high-precision fault identification and positioning for multiple sections of heterogeneous transmission cables and provide timely early warning protection.

[0010] The object of the present invention can be achieved by the following technical solution: A method for fault identification and early warning protection of an offshore wind farm transmission system based on data-driven state estimation, comprising the following steps:

[0011] S1. Build a unified global state model that adapts to the heterogeneous topology of offshore wind power and establish a state estimation model for long-distance transmission cables;

[0012] S2. Collect the measurement data at the cable end and combine it with the state estimation model of the long-distance transmission cable to obtain the state estimation;

[0013] S3. Based on the state estimate, calculate the residual and fit the error;

[0014] S4. Based on the residual and fitting error results, determine whether a fault has occurred, perform fault location classification, and output corresponding protection actions and alarm information.

[0015] Furthermore, the step S1 specifically models each cable segment separately and then achieves unified coupling through electrical constraints to construct a unified global state model. The process includes:

[0016] State vector for single-segment cable modeling is 12×1, and its measurement equation is ,in, is the state vector of a certain cable segment, is the corresponding measurement matrix;

[0017] For a three-phase system, the state vector dimension is 36×1, the measurement matrix dimension is 42×36, and the connection between each cable segment is realized by the node continuity equation: ,in, Represents the voltage between two adjacent cable segments at the connection node, Represents the current at the corresponding node, satisfying the continuity and power conservation relationship;

[0018] Finally, the modeling equations of all cable segments are spliced ​​into a unified structure, expressed as ,in represents the total amount of measurement vector at the system level, represents the global observation matrix, Represents the complete set of state variables of the entire submarine cable system.

[0019] Furthermore, the process of establishing the state estimation model of the long-distance transmission cable in step S1 includes:

[0020] Based on the idea of ​​distributed parameter modeling of cable system in differential algebraic form, state space expression is used to describe the system dynamics;

[0021] By discretizing the state space expression, the nonlinear measurement equation is obtained;

[0022] Then, the least squares estimation criterion is used, combined with the nonlinear measurement equation, to construct the optimization objective function to solve and obtain the state estimate.

[0023] Furthermore, the state space expression is specifically: , with voltage, current and other electrical quantities as state variables, where x represents the state variable, y represents the observable output variable, and t is the time variable. 、 、 are the state equation, output equation and constraint equation respectively, represents the time derivative of the state variable.

[0024] Furthermore, the nonlinear measurement equation is specifically: ,in, is the measurement vector, is a nonlinear state mapping, represents the estimated state, represents the boundary state variable, is Gaussian white noise.

[0025] Furthermore, the optimization objective function is specifically: , where W is the weight matrix and the iterative update form of the solution is: , H represents the Jacobian matrix of the measurement function to the state.

[0026] Furthermore, the calculation formula of the fitting error in step S3 is:

[0027]

[0028] in, represents the residual error on the i-th measurement channel, Indicates that the estimated state The predicted value of the i-th channel is calculated, represents the actual measurement value of the i-th channel, is the standard deviation of the i-th measurement channel.

[0029] Furthermore, the step S4 specifically includes the following steps:

[0030] S41. Detecting whether there is an abnormal measurement signal based on the fluctuation of the residual;

[0031] Detect whether a fault occurs based on whether the fitting error exceeds the preset limit;

[0032] S42: If an abnormal measurement signal is detected, locate the corresponding abnormal channel and perform elimination and fault tolerance processing;

[0033] If a fault is detected, the fault is located and classified based on the residual distribution, and the corresponding protection action and alarm information are output;

[0034] If no fault is detected, the process returns to step S2.

[0035] Furthermore, in step S41, if the fluctuation of the residual on a certain measurement channel exceeds a preset fluctuation threshold, it is determined that an abnormal measurement signal exists on the measurement channel, and the measurement channel is an abnormal channel;

[0036] If the fitting error exceeds the preset limit, a fault is determined to have occurred.

[0037] Furthermore, the specific process of performing fault location and classification based on residual distribution in step S42 is as follows:

[0038] The residual space distribution index is introduced and divided into segment fitting error percentage and phase fitting error percentage:

[0039] 1) Segment fitting error percentage:

[0040]

[0041] 2) Phase fitting error percentage:

[0042]

[0043] in, and They represent the proportion of fitting error of the s-th cable segment and the proportion of fitting error of the p-th phase channel, respectively. and They represent the residual of the ith channel in the sth segment and the residual of the ith channel in the pth phase respectively. FE is the total fitting error index of the system. The segment fitting error percentage is used to determine the segment where the cable fault occurs, and the phase fitting error percentage is used to determine the phase of the cable fault.

[0044] By drawing the residual amplitude distribution diagram of each section / phase, the fault concentration area and phase can be identified.

[0045] Compared with the prior art, the present invention has the following advantages:

[0046] The present invention first constructs a unified global state model adapted to the heterogeneous topology of offshore wind power and establishes a state estimation model for long-distance transmission cables. The collected cable-end measurement data is input into the state estimation model of the long-distance transmission cable to obtain a state estimate. Based on the state estimate, the residual and fitting error are calculated. Finally, based on the residual and fitting error results, a judgment is made as to whether a fault has occurred, fault location classification is performed, and corresponding protection actions and alarm information are output. This establishes a three-layer nested architecture of distributed parameter modeling, state solution, and residual discrimination, constructs an accurate physical model and state mapping relationship, reconstructs the system node state, and integrates residual amplitude, phase angle change, and spatial distribution characteristics to achieve multi-dimensional fault identification and section location, as well as timely early warning protection.

[0047] Taking into account that the offshore wind power transmission cable system is actually composed of multiple cable segments with different physical parameters, the present invention designs each cable segment to be modeled separately and achieves unified coupling through electrical constraints, so as to realize unified modeling of the global state of the multi-segment complex structure cable system, which can well adapt to the typical heterogeneous topology of offshore wind power.

[0048] The present invention takes into account the significant differences in physical structure and electrical parameters of multiple sections such as land cables, tidal flat cables, shallow-sea cables and deep-sea cables in the offshore wind power transmission cable system. Therefore, the sampling is based on the idea of ​​distributed parameter modeling of the cable system in differential algebraic form. First, the state space expression is used to describe the system dynamics, and then discretization is performed to obtain the nonlinear measurement equation. Finally, the least squares estimation criterion is used to establish a state estimation model, which can accurately solve the internal electrical state of the system under multi-source measurement input.

[0049] The present invention constructs a fault detection mechanism centered on fitting error. This fault detection mechanism does not rely on traditional features such as current direction or amplitude, and is suitable for converter power supply scenarios in wind power systems with low fault current amplitudes. In addition, the present invention introduces residual spatial distribution indicators (including segment fitting error percentage and phase fitting error percentage) to achieve fault location and classification based on residual distribution. By plotting the residual amplitude distribution map of each segment / phase, the fault concentration area and phase difference can be accurately identified, realizing the spatial location capability of complex faults in heterogeneous structures.

[0050] The present invention determines whether there are abnormal measurement signals based on the fluctuation of residuals, and eliminates and performs fault-tolerant processing on abnormal measurement channels with abnormal measurement signals to ensure the stability of state quantity estimation. By introducing a multi-channel redundant measurement data verification mechanism, the state estimation accuracy can be maintained in the case of single-point measurement failure or data drift, thereby improving robustness and fault-tolerant performance, and realizing a five-step mechanism of "modeling-estimation-detection-positioning-fault tolerance", which can complete high-reliability intelligent protection of long-distance complex cable systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 Schematic diagram of the method flow of the present invention;

[0052] Figure 2 This is a schematic diagram of the application process of Example 2;

[0053] Figure 3 This is a structural diagram of a multi-section heterogeneous structure offshore wind power transmission system in Example 2;

[0054] Figure 4 This is a spatial distribution diagram of system node residuals under a two-phase short circuit fault condition in Example 2;

[0055] Figure 5 This is a spatial distribution diagram of system node residuals when the measurement channel is abnormal in Example 2. DETAILED DESCRIPTION

[0056] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0057] Example 1

[0058] like Figure 1 As shown, a method for fault identification and early warning protection of an offshore wind farm transmission system based on data-driven state estimation includes the following steps:

[0059] S1. Build a unified global state model that adapts to the heterogeneous topology of offshore wind power and establish a state estimation model for long-distance transmission cables;

[0060] S2. Collect the measurement data at the cable end and combine it with the state estimation model of the long-distance transmission cable to obtain the state estimation;

[0061] S3. Based on the state estimate, calculate the residual and fit the error;

[0062] S4. Based on the residual and fitting error results, determine whether a fault has occurred, perform fault location classification, and output corresponding protection actions and alarm information.

[0063] In step S1, on the one hand, each cable segment is modeled separately and then uniformly coupled through electrical constraints to construct a unified global state model. On the other hand, a state estimation model is constructed through state space modeling and discrete processing combined with the least squares method.

[0064] In step S3, the fitting error (FE) based on the residual is introduced as the basis for fault judgment.

[0065] In step S4, on the one hand, based on the fluctuation of the residual, it is detected whether there is an abnormal measurement signal. If an abnormal measurement signal is detected, the corresponding abnormal channel is located and eliminated and fault-tolerant processing is performed;

[0066] On the other hand, whether a fault occurs is detected based on whether the fitting error exceeds the preset limit. If a fault is detected, the fault is located and classified based on the residual distribution, and the corresponding protection action and alarm information are output. Specifically, the residual spatial distribution index is introduced, which is divided into the segment fitting error percentage and the phase fitting error percentage. Among them, the segment fitting error percentage is used to achieve accurate positioning of "which section of the cable is faulty" at the spatial level, and the phase fitting error percentage is used to achieve accurate identification of "fault phase" at the electrical level.

[0067] This solution uses a model-driven state estimation method to detect cable system faults. It does not rely on traditional fault current criteria and is suitable for wind power converter systems with weak current characteristics, improving the universality and accuracy of fault identification.

[0068] This solution constructs a refined topological model that is suitable for multi-section heterogeneous cable structures. By integrating state variable solution with residual criteria, it can effectively locate the fault section and identify the phases, thereby improving the resolution capability and engineering applicability of the protection system.

[0069] This scheme introduces a multi-channel redundant measurement data verification mechanism, which can maintain the state estimation accuracy in the case of single-point measurement failure or data drift, thereby improving the robustness and fault tolerance of the system.

[0070] This solution eliminates the need for high-speed communication links and remote synchronization signals, allowing protection strategies to be deployed locally. This significantly reduces the system's reliance on communication infrastructure and enhances feasibility and operational safety in complex offshore environments. This solution, while abandoning the traditional reliance on current amplitude and communication synchronization, boasts excellent structural generalization and engineering deployment adaptability, making it suitable for practical application in next-generation high-proportion offshore wind power transmission protection scenarios.

[0071] Example 2

[0072] This embodiment applies the above technical solution, such as Figure 2 As shown, the main contents are:

[0073] Step 01. Establish a state estimation model for long-distance transmission cables: First, for a multi-section heterogeneous offshore wind power transmission cable system (such as Figure 3 As shown in the figure, a mathematical model is established to serve as the basis for the subsequent protection mechanism. The offshore wind power transmission cable system is usually composed of multiple sections such as land cables, beach cables, shallow sea cables and deep sea cables. In this embodiment, the offshore wind power transmission cable system includes a 60km deep sea cable (i.e. Figure 3 The submarine cable 2), 60km shallow sea cable (ie Figure 3 The submarine cable 1), 5km mudflat cable (ie Figure 3 The landing cable in the Figure 3 The physical structure and electrical parameters of each cable segment vary significantly. Specifically, this embodiment uses the following state-space expression to describe the system dynamics based on the distributed parameter modeling of the cable system in differential algebraic form: 1) , the model uses voltage, current and other electrical quantities as state variables, where x represents the state variable (including voltage, current, etc.), y represents the observable output variable, and t is the time variable. 、 、 are the state equation, output equation and constraint equation respectively, represents the time derivative of the state variable;

[0074] 2) Through discretization processing, the nonlinear measurement equation is obtained ,in, is the measurement vector, is a nonlinear state mapping, represents the estimated state, represents the boundary state variable, is Gaussian white noise;

[0075] 3) On this basis, the least squares estimation criterion is used to construct the optimization objective function , where W is the weight matrix; 4) The iterative update form of the solution is , where H represents the Jacobian matrix of the measurement function to the state. This state estimation model can solve the internal electrical state of the system under multi-source measurement input, providing theoretical support for subsequent fault detection and diagnosis.

[0076] Step 02. Construct a fault detection mechanism based on fitting error: Secondly, based on the state estimation, a sensitive and structure-independent fault detection mechanism must be designed to address the challenge of fault current weakening when connected to a converter-type power source. Specifically, this embodiment introduces the residual-based fitting error (FE) as the basis for fault determination. It is defined as follows:

[0077] 1) .in, represents the residual error on the i-th measurement channel, Indicates that the estimated state The predicted value of the i-th channel is calculated, represents the actual measurement value of the i-th channel, is the standard deviation of the i-th measurement channel;

[0078] 2) During normal operation, the FE value is close to zero. However, if an internal fault occurs, the state estimate and actual measurement will mismatch, causing the FE value to suddenly rise, serving as a fault trigger. This mechanism does not rely on traditional characteristics such as current direction or amplitude, and is suitable for converter-powered wind power systems with low fault current amplitudes.

[0079] Step 03. Construct segmented modeling and cascade coupling mechanism: Considering that the transmission system is actually composed of multiple cable segments with different physical parameters, each segment needs to be modeled separately and uniformly coupled through electrical constraints.

[0080] Specifically: 1) State vector for single-segment cable modeling is 12×1, and its measurement equation is as follows .in, is the state vector of a certain cable segment, is the corresponding measurement matrix;

[0081] 2) For a three-phase system, the state vector dimension is 36×1 and the measurement matrix dimension is 42×36. The inter-segment connection is realized by the node continuity equation as follows ,in Represents the voltage between two adjacent cable segments at the connection node, Represents the current at the corresponding node, satisfying the continuity and power conservation relationship;

[0082] 3) Finally, the modeling equations of all cable segments are spliced ​​into a unified structure, expressed as in represents the total amount of measurement vector at the system level, represents the global observation matrix, Represents the complete set of state variables for the entire submarine cable system. This step enables unified global state modeling of multi-section, complex cable systems, adapting to typical heterogeneous topologies of offshore wind power.

[0083] Step 4. Implement fault location and classification based on residual distribution: After a fault is detected, to achieve high-precision location and phase identification, a residual spatial distribution index is introduced, which is divided into segment fitting error percentage and phase fitting error percentage.

[0084] Specifically, this embodiment proposes the following indicators: 1) Segment fitting error percentage:

[0085]

[0086] 2) Phase fitting error percentage:

[0087]

[0088] in, and They represent the proportion of fitting error of the s-th cable segment and the proportion of fitting error of the p-th phase channel, and where represents the residual error of the i-th channel in the s-th segment and the i-th channel in the p-th phase, respectively. FE is the total system fitting error. The first segment fitting error percentage accurately locates the faulty cable segment at the spatial level, while the second phase fitting error percentage accurately identifies the faulty phase at the electrical level.

[0089] By drawing the residual amplitude distribution map of each section / phase, the fault concentration area and phase can be accurately identified, and the spatial positioning capability of complex faults in heterogeneous structures can be achieved.

[0090] This embodiment uses indicators to quantitatively determine the maximum error segment / phase and a residual amplitude distribution graph to visually identify the distribution of outliers. Specifically, the indicator is used to determine "which segment / phase is most severe for the fault," while the residual amplitude distribution graph is used to determine "where / channel the fault occurred."

[0091] Step 05. Build and deploy a robust protection application mechanism: Finally, to implement this embodiment in a complex offshore environment, it is necessary to integrate the state estimation and fault diagnosis mechanisms into a real-time and fault-tolerant protection system.

[0092] Specifically, the design is as follows: 1) The system acquires multi-channel measurement data in each sampling period, inputs it into the state estimation model, and solves the internal states of the node voltage, current, etc. in real time; 2) The fitting error index (FE) is calculated based on the estimation result. If it exceeds the preset safety threshold, the protection logic is triggered to respond to the fault; 3) At the same time, combined with the residual distribution, the fault section and fault phase are located and identified to achieve fast and accurate protection action; 4) If some measurement signals are abnormal, such as channel drift, mutual inductor saturation or communication loss of step, the system eliminates the abnormal channel through redundant judgment to ensure estimation stability. In this embodiment, Figure 4 This is a schematic diagram of the spatial distribution of system node residuals under a bidirectional short circuit fault. Figure 5 Schematic diagram of the spatial distribution of system node residuals when abnormal measurement signals occur.

[0093] In practical applications, this embodiment builds a fault identification and early warning protection system for offshore wind power transmission systems, including a data acquisition module, a state estimation module, a residual calculation and fitting error generation module, a three-branch judgment module, and an output alarm module that are communicatively connected in sequence. The data acquisition module is arranged at the cable position of the transmission system and is used to collect cable end measurement data, including voltage and current, and transmit the collected measurement data to the state estimation module.

[0094] The state estimation module is pre-set with a constructed state estimation model. After receiving the input measurement data, it can solve the output state estimation and transmit the result to the residual calculation and fitting error generation module;

[0095] The residual calculation and fitting error generation module is used to calculate the residual of each measurement channel and calculate the fitting error in combination with the state estimate;

[0096] The three-branch judgment module, on the one hand, detects whether there are abnormal measurement signals based on the fluctuation of the residuals, and eliminates and performs fault-tolerant processing on abnormal measurement channels. On the other hand, it detects whether a fault has occurred based on the fitting error. If a fault occurs, it further combines the residual distribution to complete the location and identification of the faulty section and faulty phase, and uses the output alarm module to output fast and accurate protection actions and alarm information.

[0097] This embodiment ultimately completes high-reliability intelligent protection of long-distance complex cable systems through a five-step mechanism of "modeling-estimation-detection-positioning-fault tolerance".

[0098] In summary, this scheme proposes a three-phase modeling and residual discrimination mechanism based on state estimation, integrates system distributed parameter modeling and least squares real-time state estimation method, constructs a fault detection index with fitting error as the core, and realizes sensitive identification of internal faults.

[0099] This solution establishes a local judgment system based on end-side measurement, using local voltage and current signals and physical modeling methods for fault identification. It does not rely on remote synchronous communication or current direction information and is suitable for high-resistance and low-current faults in converter-type power supply conditions.

[0100] This scheme constructs a modeling and cascade coupling mechanism for segmented heterogeneous cables. By splicing multiple segment state equations and constraining node continuity, it achieves a unified state solution for the system, and introduces a residual space distribution index to achieve segment and phase fault location and classification.

[0101] This solution designs a robust and fault-tolerant protection application framework. Combining residual threshold judgment with abnormal channel elimination strategy, it achieves highly reliable online fault detection, location and protection signal output. It is suitable for the deployment of multi-section long-distance heterogeneous submarine cable projects.

Claims

1. A method for fault identification and early warning protection of offshore wind farm transmission system based on data-driven state estimation, characterized in that: The following steps are involved: S1. Build a unified global state model that adapts to the heterogeneous topology of offshore wind power and establish a state estimation model for long-distance transmission cables; S2. Collect the measurement data at the cable end and combine it with the state estimation model of the long-distance transmission cable to obtain the state estimation; S3. Based on the state estimate, calculate the residual and fit the error; S4. Determine whether a fault has occurred and perform fault location classification based on the residual and fitting error results, and output corresponding protection actions and alarm information; Step S1 specifically models each cable segment separately and then achieves unified coupling through electrical constraints to construct a unified global state model. The process includes: State vector for single-segment cable modeling is 12×1, and its measurement equation is ,in, is the state vector of a certain cable segment, is the corresponding measurement matrix; For a three-phase system, the state vector dimension is 36×1, the measurement matrix dimension is 42×36, and the connection between each cable segment is realized by the node continuity equation: ,in, Represents the voltage between two adjacent cable segments at the connection node, Represents the current at the corresponding node, satisfying the continuity and power conservation relationship; Finally, the modeling equations of all cable segments are spliced ​​into a unified structure, expressed as ,in represents the total amount of measurement vector at the system level, represents the global observation matrix, Represents the complete set of state variables of the entire submarine cable system.

2. A method for fault identification and early warning protection of offshore wind farm transmission system based on data-driven state estimation according to claim 1, characterized in that: The process of establishing the state estimation model of the long-distance transmission cable in step S1 includes: Based on the idea of ​​distributed parameter modeling of cable system in differential algebraic form, state space expression is used to describe the system dynamics; By discretizing the state space expression, the nonlinear measurement equation is obtained; Then, the least squares estimation criterion is used, combined with the nonlinear measurement equation, to construct the optimization objective function to solve and obtain the state estimate.

3. The method for fault identification and early warning protection of offshore wind farm transmission system based on data-driven state estimation according to claim 2, characterized in that: The state space expression is specifically: , with voltage, current and other electrical quantities as state variables, where x represents the state variable, y represents the observable output variable, and t is the time variable. 、 、 are the state equation, output equation and constraint equation respectively, represents the time derivative of the state variable.

4. The method for fault identification and early warning protection of offshore wind farm transmission system based on data-driven state estimation according to claim 3, characterized in that: The nonlinear measurement equation is specifically: ,in, is the measurement vector, is a nonlinear state mapping, represents the estimated state, represents the boundary state variable, is Gaussian white noise.

5. The method for fault identification and early warning protection of offshore wind farm transmission system based on data-driven state estimation according to claim 4, characterized in that: The optimization objective function is specifically: , where W is the weight matrix and the iterative update form of the solution is: , H represents the Jacobian matrix of the measurement function to the state.

6. The method for fault identification and early warning protection of offshore wind farm transmission system based on data-driven state estimation according to claim 5, characterized in that: The calculation formula of the fitting error in step S3 is: ， in, represents the residual error on the i-th measurement channel, Indicates that the estimated state The predicted value of the i-th channel is calculated, represents the actual measurement value of the i-th channel, is the standard deviation of the i-th measurement channel.

7. The method for fault identification and early warning protection of offshore wind farm transmission system based on data-driven state estimation according to claim 6, characterized in that: The step S4 specifically includes the following steps: S41. Detecting whether there is an abnormal measurement signal based on the fluctuation of the residual; Detect whether a fault occurs based on whether the fitting error exceeds the preset limit; S42: If an abnormal measurement signal is detected, locate the corresponding abnormal channel and perform elimination and fault tolerance processing; If a fault is detected, the fault is located and classified based on the residual distribution, and the corresponding protection action and alarm information are output; If no fault is detected, the process returns to step S2.

8. The method for fault identification and early warning protection of offshore wind farm transmission system based on data-driven state estimation according to claim 7, characterized in that: In step S41, if the fluctuation of the residual on a certain measurement channel exceeds a preset fluctuation threshold, it is determined that there is an abnormal measurement signal on the measurement channel, and the measurement channel is an abnormal channel; If the fitting error exceeds the preset limit, a fault is determined to have occurred.

9. The method for fault identification and early warning protection of offshore wind farm transmission system based on data-driven state estimation according to claim 7, characterized in that: The specific process of fault location and classification based on residual distribution in step S42 is as follows: The residual space distribution index is introduced and divided into segment fitting error percentage and phase fitting error percentage: 1) Segment fitting error percentage: ， 2) Phase fitting error percentage: ， in, and They represent the proportion of fitting error of the s-th cable segment and the proportion of fitting error of the p-th phase channel, respectively. and They represent the residual of the ith channel in the sth segment and the residual of the ith channel in the pth phase respectively. FE is the total fitting error index of the system. The segment fitting error percentage is used to determine the segment where the cable fault occurs, and the phase fitting error percentage is used to determine the phase of the cable fault. By drawing the residual amplitude distribution diagram of each section / phase, the fault concentration area and phase can be identified.

Images