Fault detection method for distributed grid-connected photovoltaic power generation system based on ellipsoid estimation

Abstract

The invention discloses a fault detection method for a distributed grid-connected photovoltaic power generation system based on ellipsoid estimation. The fault detection method comprises the following steps: 1, establishing a state model and a sensor measurement model of the distributed grid-connected photovoltaic power generation system; 2, describing a state range through an ellipsoid geometric method; 3, correcting the predicted state according to the measurement data of the sensor; during correction, the center and the shape of the ellipsoid geometric method are adjusted according to the measurement data, so that the range of describing the state through the ellipsoid geometric method is more accurate; 4, fault detection is realized by judging whether the prediction ellipsoid set and the updated prediction ellipsoid set intersect or not; if the intersection of the predicted ellipsoid set and the updated predicted ellipsoid set is not an empty set, determining that no fault occurs; and if the intersection of the predicted ellipsoid set and the updated predicted ellipsoid set is empty, indicating that a fault occurs. According to the method, non-Gaussian noise and system model errors can be effectively processed, and the method is particularly suitable for fault detection in a complex dynamic environment.

Description

Technical Field

[0001] The present invention relates to the technical field of fault detection of distributed photovoltaic grid-connected power generation systems, and in particular to a fault detection method of distributed photovoltaic grid-connected power generation systems based on ellipsoid estimation. Background Art

[0002] With the continuous expansion of the scale of distributed photovoltaic grid-connected power generation systems and the widespread access to new energy, the complexity and dynamics of distributed photovoltaic grid-connected power generation systems have increased significantly.

[0003] Traditional centralized fault detection methods mainly rely on centralized data processing and monitoring, which has significant limitations when faced with the huge amount of data and complex operating environment of distributed photovoltaic grid-connected power generation systems.

[0004] Existing centralized approaches require large amounts of data to be transmitted to a central control center for processing, resulting in excessive communication burden and being susceptible to network latency and bandwidth limitations.

[0005] Moreover, due to the rapid changes in the status of distributed photovoltaic grid-connected power generation systems, especially the volatility of renewable energy power generation, traditional methods are difficult to achieve real-time and accuracy, and are unable to quickly respond to dynamic changes in distributed photovoltaic grid-connected power generation systems and identify faults in a timely manner.

[0006] More importantly, when a single point failure occurs in a centralized architecture, it may lead to the collapse of the entire system, causing large-scale shutdown of distributed photovoltaic grid-connected power generation systems, and lacks sufficient fault tolerance and robustness.

[0007] In addition, there are many types of faults in distributed photovoltaic grid-connected power generation systems, and the fault manifestations are complex and nonlinear. Traditional fault detection methods often rely on specific models and assumptions, resulting in inefficiency in handling abnormal situations.

[0008] The above existing methods cannot adapt to the ever-changing operating conditions of distributed photovoltaic grid-connected power generation systems. The dynamic changes and uncertainties of distributed photovoltaic grid-connected power generation systems pose huge challenges to fault detection.

[0009] Therefore, how to improve the real-time, robustness and accuracy of high fault detection in intelligent distributed photovoltaic grid-connected power generation systems has become a technical problem that technical personnel in this field need to solve urgently. Summary of the invention

[0010] In view of the above-mentioned defects of the prior art, the present invention provides a fault detection method for a distributed photovoltaic grid-connected power generation system based on ellipsoid estimation. The purpose of the method is to achieve distributed collaborative processing among the nodes of the distributed photovoltaic grid-connected power generation system, use the ellipsoid geometry method to describe the uncertainty of the system state, and dynamically update the state estimate of each node through a prediction-correction recursive estimation algorithm, so as to achieve low communication burden, high real-time and high robustness fault detection in a large-scale distributed photovoltaic grid-connected power generation system.

[0011] To achieve the above object, the present invention discloses a fault detection method for a distributed photovoltaic grid-connected power generation system based on ellipsoid estimation. A sensor is set at each node of the distributed photovoltaic grid-connected power generation system to form a distributed sensor structure of the distributed photovoltaic grid-connected power generation system. The following steps are performed according to the data collected by the sensor:

[0012] Step 1: Based on the dynamic characteristics and node measurement characteristics of the distributed photovoltaic grid-connected power generation system, a distributed photovoltaic grid-connected power generation system state model and a sensor measurement model are established;

[0013] Step 2, describing the range of the state by ellipsoid geometry method;

[0014] Step 3: Correct the predicted state according to the measurement data of the sensor;

[0015] In the calibration, the center and shape of the ellipsoid geometry method are adjusted according to the measurement data so that the range of the state described by the ellipsoid geometry method is more accurate;

[0016] Step 4: Fault detection is performed by determining whether the predicted ellipsoid set and the updated predicted ellipsoid set have an intersection; if the intersection of the predicted ellipsoid set and the updated predicted ellipsoid set is not an empty set, no fault has occurred; if the intersection of the predicted ellipsoid set and the updated predicted ellipsoid set is empty, a fault has occurred.

[0017] Preferably, in step 1, the distributed photovoltaic grid-connected power generation system model is constructed as follows:

[0018] The discrete voltage state at the common coupling point (PCC) of the distributed photovoltaic grid-connected power generation system is the measured value to be observed;

[0019] Among them, Y k represents the voltage amplitude, ω is the angular frequency, represents the initial phase, k is the kth sampling, and T is the sampling period;

[0020] There will be a voltage deviation between the two sampling intervals, namely Δy k =Y k+1 -Y k .

[0021] Define the state variables as:

[0022] Then the state estimation system: k+1 =α k χ k +β k γ k ;

[0023]

[0024] Among them, χ k is the state variable of the system;

[0025] γ k represents process noise, which is unknown but bounded;

[0026] α k and β k Represents a known real-time changing matrix with appropriate dimensions

[0027] The measurement model of a distributed photovoltaic grid-connected power generation system with N sensors is as follows:

[0028]

[0029] in, θ i,k =2+0.1i,ρ i,k Indicates the detection output;

[0030] ο i,k represents the measurement noise, which is unknown but bounded;

[0031] ι i,k represents the fault signal of the i-th sensor;

[0032] G i,k Represents a known real-time changing matrix with appropriate dimensions.

[0033] More preferably, in step 2, it is assumed that the noise and the fault signal are a known ellipsoid set, as follows:

[0034] γ k ∈ε(0,T k );

[0035] ο i,k ∈ε(0,S i,k );

[0036] ι i,k ∈ε(0,H i,k );

[0037] The ellipsoid set is defined as: ε(a,p)={χ|(χ-a)T P -1 (χ-a)≤1};

[0038] Among them, a represents the center of the ellipsoid, and P is a matrix representing the shape of the ellipsoid;

[0039] T k , S i,k and H i,k (i∈{1,2,…,N}) is a known matrix with common dimensions.

[0040] More preferably, step 3 is performed recursively, using the measurement data measured in real time to adjust the center and shape of the range describing the state by the ellipsoid geometry method, as follows:

[0041] Step 3.1: Set the prediction error ζ i,k+1|k Defined as:

[0042]

[0043] Measurement update error ζ i,k+1|k+1 Defined as:

[0044]

[0045] Step 3.2: Using the recursive linear matrix inequality (RLMI), the filtering errors are obtained to satisfy the P-based i,k+1|k Constraints and P-based i,k+1|k+1 Sufficient conditions for constraints;

[0046] Step 3.3, by solving the convex optimization problem, two optimized ellipsoid sets are obtained recursively;

[0047] Step 3.4: Use the intersection of the two ellipsoid sets to determine whether a fault has occurred, by minimizing P i,k+1|k and P i,k+1|k+1 Get the optimal ellipsoid set, as follows:

[0048] Step 3.4.1. Initialize the ellipsoid Recursive times T, let k = 0;

[0049] Step 3.4.2: When k>T, loop through the following steps:

[0050] Step 3.4.2.1, by solving Get the matrix Q i,k+1|k ;

[0051] Among them, τ 1，i ≥0,τ 2,i ≥0 is a scalar sequence;

[0052] Step 3.4.2.2: Determined by the S-procedure lemma Solving for the Ellipsoid the center of;

[0053] Step 3.4.2.3, by solving Get the matrix Q i,k+1|k+1 ;

[0054] in, and is a positive scalar sequence, C ij,k+1 ,B ij,k+1 is a real matrix sequence;

[0055] Step 3.4.2.4: Use Finsler’s lemma to determine that there is a set d2 that satisfies the formula χ k+1 = x i,k+1|k +Q i,k+1|k d2 Solving for the Ellipsoid The center of the ellipsoid, where d2 is an auxiliary variable used to describe the deviation between the internal point of the ellipsoid and the center position, d2 T d2≤1.

[0056] More preferably, if there is a fault, the updated estimated ellipsoid set At the current moment, k+1 will change as follows:

[0057] When adjusting the center and shape of the range described by the ellipsoid geometry method, the matrix update rule is:

[0058]

[0059] The predicted ellipsoid set is:

[0060] When adjusting the center and shape of the range described by the ellipsoid geometry method, the matrix update rule is:

[0061]

[0062] Then the updated prediction ellipsoid set is:

[0063] Beneficial effects of the present invention:

[0064] The present invention is based on ellipsoid estimation and describes the uncertainty of the state of distributed photovoltaic grid-connected power generation system in a geometric way. It can effectively handle non-Gaussian noise and system model errors and has strong robustness. It is particularly suitable for fault detection of distributed photovoltaic grid-connected power generation system in complex dynamic environments.

[0065] The present invention aims at the distributed sensor structure of the distributed photovoltaic grid-connected power generation system, so that each node can independently complete state estimation by using local data and neighborhood communication, without the need for centralized processing, which greatly reduces the communication burden. At the same time, real-time performance is achieved through recursive updating, which can quickly respond to state changes of the distributed photovoltaic grid-connected power generation system and accurately locate faults.

[0066] The present invention has low computational complexity and strong adaptability, and can be extended to large-scale distributed photovoltaic grid-connected power generation systems containing a large number of sensor nodes. Its flexibility ensures that adding or removing nodes will not affect the overall operation, and is suitable for the dynamic perception and fault location needs of distributed photovoltaic grid-connected power generation systems after the access of new energy.

[0067] The concept, specific structure and technical effects of the present invention will be further described below in conjunction with the accompanying drawings to fully understand the purpose, characteristics and effects of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 A structural diagram of a distributed sensor in a power grid according to an embodiment of the present invention is shown.

[0069] Figure 2 A flow chart of an algorithm for finding an optimal ellipsoid set in one embodiment of the present invention is shown. DETAILED DESCRIPTION

[0070] Example

[0071] like Figure 1 As shown, a fault detection method for a distributed photovoltaic grid-connected power generation system based on ellipsoid estimation is provided; it is characterized in that a sensor is set at each node of the distributed photovoltaic grid-connected power generation system to form a distributed sensor structure of the distributed photovoltaic grid-connected power generation system, and the following steps are performed according to the data collected by the sensor:

[0072] Step 1: Based on the dynamic characteristics of the power grid and the node measurement characteristics, a power grid state model and a sensor measurement model are established;

[0073] Step 2, describing the range of the state by ellipsoid geometry method;

[0074] Step 3: Correct the predicted state according to the measurement data of the sensor;

[0075] In the calibration, the center and shape of the ellipsoid geometry method are adjusted according to the measurement data, so that the range of the state described by the ellipsoid geometry method is more accurate;

[0076] Step 4: Fault detection is performed by determining whether the predicted ellipsoid set and the updated predicted ellipsoid set have an intersection; if the intersection of the predicted ellipsoid set and the updated predicted ellipsoid set is not an empty set, no fault has occurred; if the intersection of the predicted ellipsoid set and the updated predicted ellipsoid set is empty, a fault has occurred.

[0077] The present invention uses distributed ellipsoid estimation to describe the uncertainty of the state of a distributed photovoltaic grid-connected power generation system. Through the state estimation of local nodes and the prediction-correction mechanism of the ellipsoid estimation, the state estimation is updated in real time in combination with measurement data, and faults in the distributed photovoltaic grid-connected power generation system are detected in real time. Specifically, the uncertainty of the state of the distributed photovoltaic grid-connected power generation system is described by using ellipsoid geometry. Compared with the estimation method based on probability distribution, the ellipsoid estimation can better handle non-Gaussian noise and system model errors, and provide a more robust and accurate state estimation.

[0078] In some embodiments, in step 1, a distributed photovoltaic grid-connected power generation system model is constructed as follows:

[0079] The discrete voltage state at the common coupling point (PCC) of the distributed photovoltaic grid-connected power generation system is the measured value to be observed;

[0080] Among them, Y k represents the voltage amplitude, ω is the angular frequency, represents the initial phase, k is the kth sampling, and T is the sampling period;

[0081] There will be a voltage deviation between the two sampling intervals, namely Δy k =Y k+1 -Y k .

[0082] Define the state variables as:

[0083] Then the state estimation system: k+1 =α k χ k +β k γ k ;

[0084]

[0085] Among them, χ k is the state variable of the system;

[0086] γ k represents process noise, which is unknown but bounded;

[0087] α k and β k Represents a known real-time changing matrix with appropriate dimensions

[0088] The measurement model of a distributed photovoltaic grid-connected power generation system with N sensors is as follows:

[0089]

[0090] in, θ i,k =2+0.1i,ρ i,k Indicates the detection output;

[0091] ο i,k represents the measurement noise, which is unknown but bounded;

[0092] ι i,k represents the fault signal of the i-th sensor;

[0093] G i,k Represents a known real-time changing matrix with appropriate dimensions.

[0094] The present invention describes the evolution of the state of the distributed photovoltaic grid-connected power generation system over time through a distributed photovoltaic grid-connected power generation system state model, while taking into account the influence of process noise and measurement noise. The sensor measurement model reflects the relationship between the sensor output and the state of the distributed photovoltaic grid-connected power generation system, allowing independent state estimation at each node to ensure the effectiveness of the distributed architecture.

[0095] In some embodiments, in step 2, it is assumed that the noise and fault signal are known ellipsoid sets, as follows:

[0096] γ k ∈ε(0,T k );

[0097] ο i,k ∈ε(0,S i,k );

[0098] ι i,k ∈ε(0,H i,k );

[0099] The ellipsoid set is defined as: ε(a,p)={χ|(χ-a) T P -1 (χ-a)≤1};

[0100] Among them, a represents the center of the ellipsoid, and P is a matrix representing the shape of the ellipsoid;

[0101] T k , S i,k and H i,k (i∈{1,2,…,N}) is a known matrix with a common dimension. Figure 2 As shown, in some embodiments, step 3 is performed recursively, using the measurement data measured in real time to adjust the center and shape of the range describing the state by the ellipsoid geometry method, as follows:

[0102] Step 3.1: Set the prediction error ζ i,k+1|k Defined as:

[0103]

[0104] Measurement update error ζ i,k+1|k+1 Defined as:

[0105]

[0106] Step 3.2: Using the recursive linear matrix inequality (RLMI), the filtering errors are obtained to satisfy the P-based i,k+1|k Constraints and P-based i,k+1|k+1 Sufficient conditions for constraints;

[0107] Step 3.3, by solving the convex optimization problem, two optimized ellipsoid sets are obtained recursively;

[0108] Step 3.4: Use the intersection of the two ellipsoid sets to determine whether a fault has occurred, by minimizing P i,k+1|k and P i,k+1|k+1 Get the optimal ellipsoid set, as follows:

[0109] Step 3.4.1. Initialize the ellipsoid Recursive times T, let k = 0;

[0110] Step 3.4.2: When k>T, loop through the following steps:

[0111] Step 3.4.2.1, by solving Get the matrix Q i,k+1|k ;

[0112] Among them, τ 1，i ≥0,τ 2,i ≥0 is a scalar sequence;

[0113] Step 3.4.2.2: Determined by the S-procedure lemma Solving for the Ellipsoid the center of;

[0114] Step 3.4.2.3, by solving Get the matrix Q i,k+1|k+1 ;

[0115] in, and is a positive scalar sequence, C ij,k+1 ,B ij,k+1 is a real matrix sequence;

[0116] Step 3.4.2.4: Use Finsler’s lemma to determine that there is a set d2 that satisfies the formula χ k+1 = x i,k+1|k +Qi,k+1 | k d2 Solving for the Ellipsoid The center of the ellipsoid, where d2 is an auxiliary variable used to describe the deviation between the internal point of the ellipsoid and the center position, d2 T d2≤1.

[0117] In some embodiments, if there is a fault, the updated estimated ellipsoid set At the current moment, k+1 will change as follows:

[0118] When adjusting the center and shape of the range that describes the state through the ellipsoid geometry method, the matrix update rule is:

[0119]

[0120] The predicted ellipsoid set is:

[0121] When adjusting the center and shape of the range that describes the state through the ellipsoid geometry method, the matrix update rule is:

[0122]

[0123] Then the updated prediction ellipsoid set is:

[0124] The present invention combines the predicted value through the above technical means and the measured value ρ i,k .

[0125] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A fault detection method for a distributed photovoltaic grid-connected power generation system based on ellipsoid estimation; characterized in that: Sensors are set at each node of the distributed photovoltaic grid-connected power generation system to form a distributed sensor structure of the distributed photovoltaic grid-connected power generation system. The following steps are performed according to the data collected by the sensors: Step 1: Based on the dynamic characteristics and node measurement characteristics of the distributed photovoltaic grid-connected power generation system, a distributed photovoltaic grid-connected power generation system state model and a sensor measurement model are established; Step 2, describing the range of the state by ellipsoid geometry method; Step 3: Correct the predicted state according to the measurement data of the sensor; In the calibration, the center and shape of the ellipsoid geometry method are adjusted according to the measurement data so that the range of the state described by the ellipsoid geometry method is more accurate; Step 4: Fault detection is performed by determining whether the predicted ellipsoid set and the updated predicted ellipsoid set have an intersection; if the intersection of the predicted ellipsoid set and the updated predicted ellipsoid set is not an empty set, no fault has occurred; if the intersection of the predicted ellipsoid set and the updated predicted ellipsoid set is empty, a fault has occurred.

2. The fault detection method for distributed photovoltaic grid-connected power generation system based on ellipsoid estimation according to claim 1 is characterized in that: In step 1, the distributed photovoltaic grid-connected power generation system model is constructed as follows: The discrete voltage state at the common coupling point (PCC) of the distributed photovoltaic grid-connected power generation system is the measured value to be observed; Among them, Y k represents the voltage amplitude, ω is the angular frequency, represents the initial phase, k is the kth sampling, and T is the sampling period; There will be a voltage deviation between the two sampling intervals, namely Δy k =Y k+1 -Y k . Define the state variables as: Then the state estimation system: k+1 =α k χ k +β k γ k ; Among them, χ k is the state variable of the system; γ k represents process noise, which is unknown but bounded; α k and β k Represents a known real-time changing matrix with appropriate dimensions The measurement model of a distributed photovoltaic grid-connected power generation system with N sensors is as follows: in, θ i,k =2+0.1i,ρ i,k Indicates the detection output; ο i,k represents the measurement noise, which is unknown but bounded; ι i,k represents the fault signal of the i-th sensor; G i,k Represents a known real-time changing matrix with appropriate dimensions.

3. The fault detection method for distributed photovoltaic grid-connected power generation system based on ellipsoid estimation according to claim 2 is characterized in that: In step 2, it is assumed that the noise and fault signals are known ellipsoid sets, as follows: γk∈ε(0,T k ); the i,k ∈ε(0,S i,k ); i i,k ∈ε(0,H i,k ); The ellipsoid set is defined as: ε(a,p)={χ|(χ-a) T P -1 (χ-a)≤1}; Among them, a represents the center of the ellipsoid, and P is a matrix representing the shape of the ellipsoid; T k , S i,k and H i,k (i∈{1,2,…,N}) is a known matrix with common dimensions.

4. The fault detection method for distributed photovoltaic grid-connected power generation system based on ellipsoid estimation according to claim 3 is characterized in that: Step 3 is performed recursively, using the measurement data measured in real time to adjust the center and shape of the range describing the state by the ellipsoid geometry method, as follows: Step 3.1: Set the prediction error ζ i,k+1k Defined as: Measurement update error ζ i,k+1k+1 Defined as: Step 3.2: Using the recursive linear matrix inequality (RLMI), the filtering errors are obtained to satisfy the P-based i,k+1k Constraints and P-based i,k+1k+1 Sufficient conditions for constraints; Step 3.3, by solving the convex optimization problem, two optimized ellipsoid sets are obtained recursively; Step 3.4: Use the intersection of the two ellipsoid sets to determine whether a fault has occurred, by minimizing P i,k+1k and P i,k+1k+1 Get the optimal ellipsoid set, as follows: Step 3.4.

1. Initialize the ellipsoid Recursive times T, let k = 0; Step 3.4.2: When k>T, loop through the following steps: Step 3.4.2.1, by solving Get the matrix Q i,k+1k ; Among them, τ 1，i ≥0,τ 2,i ≥0 is a scalar sequence; Step 3.4.2.2: Determined by the S-procedure lemma Solving for the Ellipsoid the center of; Step 3.4.2.3, by solving Get the matrix Q i,k+1k+1 ; in, and is a positive scalar sequence, C ij,k+1 ,B ij,k+1 is a real matrix sequence; Step 3.4.2.4: Use Finsler’s lemma to determine that there is a set d2 that satisfies the formula χ k+1 = x i,k+1|k +Q i,k+1|k d2 Solving for the Ellipsoid The center of the ellipsoid, where d2 is an auxiliary variable used to describe the deviation between the internal point of the ellipsoid and the center position, d2 T d2≤1.

5. The fault detection method for distributed photovoltaic grid-connected power generation system based on ellipsoid estimation according to claim 4 is characterized in that: If there is a fault, the updated estimated ellipsoid set At the current moment, k+1 will change as follows: When adjusting the center and shape of the range described by the ellipsoid geometry method, the matrix update rule is: P i,k+1k =a k P i,kk a k T +b k T k b k T ; The predicted ellipsoid set is: When adjusting the center and shape of the range described by the ellipsoid geometry method, the matrix update rule is: Then the updated prediction ellipsoid set is:

Images