Distributed photovoltaic power station fault analysis method of quantum K-Means algorithm model

Abstract

The invention discloses a distributed photovoltaic power station fault analysis method based on a quantum K-Means algorithm model, and the method comprises the steps: collecting and processing the operation data information of a photovoltaic power station through a data collection module, and introducing a block chain technology to improve the safety of the data information; according to the method, fault features are extracted through a recurrent neural network deep learning model, feature data extracted by a recurrent neural network are encoded into quantum states through quantum calculation, and the encoded quantum states are clustered by using a quantum K-Means algorithm, so that the clustering process can be significantly accelerated, the calculation efficiency of fault analysis can be improved, and real-time fault analysis can be realized; the running data of the photovoltaic power station are deeply analyzed through the quantum K-Means algorithm model, and the potential fault trend and risk can be predicted.

Description

Technical Field

[0001] The present invention relates to the technical field of quantum K-Means algorithms, and more specifically to a method for fault analysis of distributed photovoltaic power stations based on a quantum K-Means algorithm model. Background Art

[0002] A distributed photovoltaic power station, also known as a decentralized power generation or distributed energy supply system, is a relatively small photovoltaic power generation and supply system configured at the user's site or near the power consumption site. This type of power generation system typically utilizes decentralized resources, has a small installed capacity, is arranged near the user, and is generally connected to a power grid with a voltage level below 35 kV or even lower.

[0003] Distributed photovoltaic power stations are small in scale, remote and scattered in location, have complex and diverse terrains, and the output power of photovoltaic modules is relatively small. A large number of photovoltaic modules need to be connected in series and parallel to form strings and arrays, resulting in a large number of power station equipment. It is prone to complex operation and maintenance, and there are many inconsistencies in current, voltage, power, etc. This not only affects the power generation efficiency of the photovoltaic power station but also may pose potential safety hazards to the power generation system. Therefore, how to achieve efficient and accurate analysis of distributed photovoltaic power station faults has become an urgent problem to be solved. Summary of the Invention

[0004] Aiming at the above technical deficiencies, the present invention discloses a method for fault analysis of distributed photovoltaic power stations based on a quantum K-Means algorithm model, which can significantly accelerate the clustering process, improve the computational efficiency of fault analysis, achieve real-time fault analysis, and enhance data security. Through in-depth analysis of the operation data of the photovoltaic power station by the quantum K-Means algorithm model, it is possible to predict potential fault trends and risks.

[0005] To achieve the above technical effects, the present invention adopts the following technical solutions: A method for fault analysis of distributed photovoltaic power stations based on a quantum K-Means algorithm model, which includes the following steps: Step 1: Collect fault data information of the distributed photovoltaic power station through a data acquisition module. The data information includes at least power output, temperature, humidity, light intensity, voltage, ripple, and current. The data acquisition module includes a sensor module, a data preprocessing module, and a data conversion module to achieve real-time acquisition of the data information. Step 2: Construct a blockchain network model, transfer the fault data information of the distributed photovoltaic power station through the blockchain network model, and package the processed data into a transaction for data uploading to the chain to achieve the security of the data information. The blockchain network model includes a blockchain layer, a consensus mechanism, a data layer, and a network layer. Step 3: Extract the fault data information of the distributed photovoltaic power station through the feature extraction model, and automatically learn the deep feature representation of the data through the deep learning model to extract higher-level abstract features from the original data; Step 4: Diagnose and analyze the fault data information of the distributed photovoltaic power station through the quantum K-Means algorithm clustering model. Through the state superposition and quantum entanglement characteristics of quantum bits, efficient search and clustering of multi-dimensional data are realized. The quantum K-Means algorithm clustering model includes a quantum encoding module, a quantum measurement module, and a clustering data feature calculation module. The output end of the quantum encoding module is connected to the input end of the quantum measurement module, and the output end of the quantum measurement module is connected to the input end of the clustering data feature calculation module; Step 5: Establish a fault prediction model based on the clustering results, and analyze and predict the future faults of the power station by receiving the real-time data information of the distributed photovoltaic power station.

[0006] As a further technical solution of the present invention, the sensor module includes power output, temperature, humidity, light intensity, voltage, ripple, and current sensors. The data preprocessing module uses a first-order low-pass filter to eliminate noise and interference, and the data conversion module performs data conversion through the ADC analog-to-digital conversion formula.

[0007] As a further technical solution of the present invention, the construction of the blockchain network model includes the following steps: Step 1: Select the Hyperledger blockchain platform and perform network configuration according to the actual needs of the distributed photovoltaic power station; Step 2: Set multiple nodes in the photovoltaic power station, including data collection nodes, analysis nodes, and verification nodes; Step 3: According to the network scale and performance requirements, select the proof-of-work consensus mechanism to ensure the security and reliability of the data in the network.

[0008] As a further technical solution of the present invention, the quantum encoding module uses quantum ground state encoding to convert the fault data information of the distributed photovoltaic power station into a quantum state. The quantum measurement module uses a controlled swap gate to calculate the similarity between the data point and the clustering center. The clustering data feature calculation module uses a quantum minimum search algorithm to find the most similar clustering center point to achieve efficient identification of different types of fault data information of the distributed photovoltaic power station. As a further technical solution of the present invention, the data collection module realizes the collection of data information including the following steps: Step 1: Eliminate noise and interference through a first-order low-pass filter, and the recurrence formula in the discrete time domain can be expressed as: In Equation (1), x[n] is the input signal, y[n] is the filtered output signal, and α is the coefficient of the filter, which determines the cut-off frequency and attenuation characteristics of the filter; Step 2: Implement data information conversion through the ADC analog-to-digital conversion formula, convert the analog voltage into a digital code, and the output information content is represented by the formula: In Equation (2), where Vmin and Vmax are the input voltage ranges of the ADC, and N is the number of bits of the ADC. After Equation (2), the conversion of the analog voltage into a digital code can be achieved, and further the processing and calculation of the analog signal can be realized. As a further technical solution of the present invention, the deep learning model in the feature extraction model is a recurrent neural network deep learning model.

[0009] As a further technical solution of the present invention, the method for the recurrent neural network deep learning model to complete feature extraction is as follows: The basic structure of the recurrent neural network includes an input layer, a hidden layer, and an output layer. At each time step (t), the RNN receives an input and updates its hidden state and then generates an output The identified function of the data information of the device in the industrial Internet environment after learning is: In Equation (3), where represents the weight matrix from the hidden layer to the output layer, represents the bias variable of the output layer, represents the weight matrix of the output layer, represents the hidden state, and the output of the hidden state is: In Equation (4), represents the weight matrix input to the hidden layer, represents the weight matrix from the hidden layer to the hidden layer, represents the bias vector of the hidden layer, represents the first detection error value, and the loss function is calculated through the mean square error loss as: In Equation (5), N represents the number of samples, represents the true value of the i-th sample, represents the predicted value of the deep learning model for the i-th sample, and the weight matrix is further updated through the optimization algorithm as: In Equation (6), represents the original point weight matrix, represents the learning rate, represents the total value in the operation path, represents the prediction threshold, represents the gradient of the loss function with respect to the weight matrix. As a further technical solution of the present invention, the method for the quantum K-Means algorithm model to implement the diagnosis and analysis of the fault data information of the distributed photovoltaic power station is as follows: Input the data set where n is the scale of the data set, the data set is divided into k categories, and the cluster centers are Then the quantum states of the data points and the cluster centers are: In formula (7), represents the j-th eigenvalue of the i-th data point, d represents the number of eigenvalues. In formula (8), represents the j-th eigenvalue of the i-th data point, k represents the number of data set categories. Calculate the similarity between any data point and the k cluster centers by the controlled swap gate Control-Swap: In formula (9), represents the cosine value between and c. Taking the quantum state as the input of the phase estimation algorithm, we can get: The similarity between is stored on the quantum bit The smaller is, the greater the similarity between the two. The steps to find the minimum value from the quantum state using the quantum minimum finding algorithm are as follows: (1) Prepare the initial value of with the quantum state (2) as the input, b as the control input, and use the Grover algorithm to find (3) If then use

[0010] and go back to step (1), repeat ​Different from conventional technologies, the present invention discloses a method for fault analysis of a distributed photovoltaic power station based on a quantum K-Means algorithm model. The method collects and processes the operation data information of the photovoltaic power station through a data acquisition module, and introduces blockchain technology to improve the security of the data information. The method extracts fault features through a recurrent neural network deep learning model, encodes the feature data extracted by the recurrent neural network into quantum states through quantum computing, and uses the quantum K-Means algorithm to cluster the encoded quantum states, which can significantly accelerate the clustering process, improve the computational efficiency of fault analysis, and achieve real-time fault analysis. Through in-depth analysis of the operation data of the photovoltaic power station by the quantum K-Means algorithm model, it is possible to predict potential fault trends and risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings, where: Figure 1 It is a schematic flow chart of the method of the present invention; Figure 2 It is a method for collecting data by the data acquisition module in the present invention; Figure 3 It is a schematic structural diagram of the recurrent neural network deep learning model in the present invention; Figure 4 It is a schematic working flow chart of the recurrent neural network deep learning model in the present invention; Figure 5 It is a schematic diagram of the principle of the control swap gate in the present invention; Figure 6 It is a schematic working flow chart of the quantum K-Means algorithm in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0012] The following describes the preferred embodiments of the present invention with reference to the drawings. It should be understood that the embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0013] As Figures 1-6 shown, a method for fault analysis of a distributed photovoltaic power station based on a quantum K-Means algorithm model includes the following steps: Step 1: Collect the fault data information of the distributed photovoltaic power station through a data acquisition module. The data information at least includes power output, temperature, humidity, light intensity, voltage, ripple, and current. The data acquisition module includes a sensor module, a data preprocessing module, and a data conversion module to achieve real-time collection of the data information; Step 2: Construct a blockchain network model, transmit the fault data information of the distributed photovoltaic power station through the blockchain network model, package the processed data into a transaction for data on-chain, and realize the security of data information. The blockchain network model includes a blockchain layer, a consensus mechanism, a data layer, and a network layer; Step 3: Extract the fault data information of the distributed photovoltaic power station through a feature extraction model, and automatically learn the deep feature representation of the data through a deep learning model to extract higher-level abstract features from the original data; Step 4: Diagnose and analyze the fault data information of the distributed photovoltaic power station through a quantum K-Means algorithm clustering model. Through the state superposition and quantum entanglement characteristics of quantum bits, efficient search and clustering of multi-dimensional data are realized. The quantum K-Means algorithm clustering model includes a quantum encoding module, a quantum measurement module, and a clustering data feature calculation module. The output end of the quantum encoding module is connected to the input end of the quantum measurement module, and the output end of the quantum measurement module is connected to the input end of the clustering data feature calculation module; The core task of the quantum encoding module is to map the classical fault data information of the distributed photovoltaic power station to the quantum state space. Classical data is usually multi-dimensional, containing multiple features such as voltage, current, power, etc., while quantum states have characteristics such as superposition and entanglement, and can efficiently represent a large amount of information on a limited number of quantum bits. Encoding method selection: Common encoding methods include ground state encoding, angle encoding, etc. Ground state encoding directly maps classical data to the ground state combination of quantum bits; angle encoding converts data features into the rotation angles of quantum bits. Different encoding methods are suitable for different types of data and problem scenarios. When allocating quantum bit resources, the number of quantum bits is reasonably allocated according to the dimension and scale of the data. The higher the data dimension, usually more quantum bits are required to accurately represent the data information, but the increase in the number of quantum bits will also bring higher computational complexity and resource consumption. Taking angle encoding as an example, assume we have an n-dimensional classical data vector For each eigenvalue , it can be encoded as the rotation angle of a quantum bit. Usually can be related to through a certain functional relationship, such as , where f is a suitable function, such as a linear function or a non-linear function. Then use a single quantum bit rotation gate to operate on the initial quantum state to obtain the encoded quantum state In this way, the classical data vector is encoded into a quantum state, and subsequent calculations can be performed using the characteristics of the quantum state. The quantum measurement module is mainly used to calculate the similarity between the quantum states of data points and the quantum states of each cluster center. In the quantum K-Means algorithm, the calculation of similarity is based on the inner product of quantum states or other quantum metrics. By calculating the similarity, it is possible to determine the cluster to which each data point most likely belongs. During quantum gate operations, quantum gates (such as the controlled-swap gate, Hadamard gate, etc.) are used to operate on the encoded quantum state to achieve the calculation of similarity. These quantum gate operations can make full use of the superposition and entanglement characteristics of quantum states to improve the calculation efficiency. After measuring the quantum state, a series of measurement results will be obtained. Statistical analysis needs to be performed on these measurement results to determine the similarity between the data points and the cluster centers. The measurement results are usually probability distributions, and through the analysis of the probability distributions, the most likely cluster membership can be obtained. Take the calculation of similarity using the controlled-swap gate as an example. Suppose there are two quantum states , which represent the quantum states of the data point and the cluster center respectively. First, these two quantum states and an auxiliary qubit are combined to form a composite quantum state . Then, the Hadamard gate is applied to the auxiliary qubit to obtain . Next, the controlled-swap gate is used. When the auxiliary qubit is , the states of are swapped. Finally, the Hadamard gate is applied to the auxiliary qubit again and a measurement is performed. The probability P(0) of measuring the auxiliary qubit as is related to the square of the inner product of the two quantum states , that is . By measuring the probability (P(0)), the similarity between the two quantum states can be obtained. The clustering data feature calculation module updates the centers of each cluster according to the cluster membership results of the data points obtained by the quantum measurement module. The update of the cluster centers is one of the core steps of the K-Means algorithm. By continuously iteratively updating the cluster centers, the clustering results gradually converge. When this module calculates the cluster centers, it usually needs to combine the results of classical computing and quantum computing. The quantum measurement module provides the cluster membership information of the data points, while the update of the cluster centers needs to be calculated on a classical computer because it involves operations such as summing and averaging a large amount of data. After each iterative update of the cluster centers, it is necessary to determine whether the algorithm converges. The criterion for convergence determination is usually that the change in the cluster centers is less than a preset threshold, or the number of iterations reaches the maximum allowed number. Suppose that at the t-th iteration, the centers of each cluster have been obtained, where k is the number of clusters. The quantum measurement module determines each data point The cluster ji to which it belongs. Then, for each cluster j, update its center The formula is: where is the number of data points belonging to cluster j. By continuously repeating this process until the convergence condition is met, a stable clustering result is finally obtained.

[0014] Step 5: Establish a fault prediction model based on the clustering result. By receiving the real-time data information of the distributed photovoltaic power station, analyze and predict the future faults of the power station. In Step 1, the sensor module includes power output, temperature, humidity, light intensity, voltage, ripple, and current sensors. The data preprocessing module uses a first-order low-pass filter to eliminate noise and interference. The data conversion module performs data conversion through the ADC analog-to-digital conversion formula. In the present invention, the sensor module is arranged in the photovoltaic array, inverter, and battery energy storage system to collect the operation data and status information of the distributed photovoltaic power station. The data acquisition module realizes the acquisition of data information, including the following steps: Step 1: Eliminate noise and interference through a first-order low-pass filter. The recurrence formula in the discrete-time domain can be expressed as: In formula (1), x[n] is the input signal, y[n] is the filtered output signal, and α is the coefficient of the filter, which determines the cut-off frequency and attenuation characteristics of the filter; Step 2: Realize data information conversion through the ADC analog-to-digital conversion formula, convert the analog voltage into a digital code, and the output information content is expressed by the formula: In formula (2), where Vmin and Vmax are the input voltage ranges of the ADC, and N is the number of bits of the ADC. After passing through formula (2), the analog voltage can be converted into a digital code, thereby realizing the processing and calculation of analog signals. The data acquisition module plays a crucial role in the fault analysis of the distributed photovoltaic power station, can monitor the operation status of each device and component in the distributed photovoltaic power station in real time, ensure that faults or abnormalities of the devices can be detected in time, and provide comprehensive and accurate data support for subsequent fault analysis. The sensor module real-time collects the power output, temperature, humidity, light intensity, voltage, ripple, and current data during the operation of the distributed photovoltaic power station, transmits the collected data information of the distributed photovoltaic power station to the data preprocessing module in real time, eliminates the noise and interference of the operation of the distributed photovoltaic power station through a first-order low-pass filter to reduce the influence of the environment on the data, eliminates the influence of different dimensions and orders of magnitude on the analysis result, and transmits the processed data information of the distributed photovoltaic power station to the data conversion module, and realizes the conversion of the data information of the distributed photovoltaic power station from analog voltage to digital code through the ADC analog-to-digital conversion formula.

[0015] In Step 2, the construction of the blockchain network model includes the following steps: Step 1: Select the Hyperledger blockchain platform and perform network configuration according to the actual requirements of the distributed photovoltaic power station; Step 2: Set multiple nodes in the photovoltaic power station, including data acquisition nodes, analysis nodes, and verification nodes; Step 3: Select the proof-of-work consensus mechanism according to the network scale and performance requirements to ensure the security and reliability of the data in the network.

[0016] With the wide application of distributed photovoltaic power generation systems, how to effectively detect, diagnose, and predict faults has become a research hotspot. With the continuous development of quantum computing and blockchain technology, these two technologies provide new possibilities for fault analysis in distributed photovoltaic power stations. The quantum K-Means algorithm model can utilize the parallelism and efficiency of quantum computing to accelerate the process of data processing and pattern recognition; while blockchain technology can provide a decentralized, tamper-proof, and transparently traceable data storage and sharing mechanism. After the data of the distributed photovoltaic power station is preprocessed, it is uploaded to each data block in the blockchain network through nodes. Each data block can store the data of the photovoltaic power station, realizing the distributed storage and backup of data, and improving the reliability and availability of data. Each data block contains a certain number of data records. Once the data block is added to the blockchain, the fault data and related operations will be recorded on the blockchain and verified and encrypted to prevent the data from being tampered with or forged. All nodes can access and obtain this data. The encryption technology and distributed storage mechanism of the blockchain have unique decentralized, tamper-proof, and transparent characteristics, which can ensure the security and integrity of the data information of the distributed photovoltaic power station during the transmission and storage processes. All the data recorded on the blockchain has timestamp and authentication information, and faults can be accurately traced to the source and the responsible party can be determined, realizing the improvement of the efficiency and accuracy of fault handling and reducing liability disputes. In Step 3, the deep learning model in the feature extraction model is a recurrent neural network deep learning model. The method for the recurrent neural network deep learning model to complete feature extraction is: The basic structure of the recurrent neural network includes an input layer, a hidden layer, and an output layer. At each time step (t), the RNN receives an input and updates its hidden state and then generates an output Then the learned function for identifying the data information of devices in the industrial Internet environment is: In formula (3), where represents the weight matrix from the hidden layer to the output layer, Represents the bias variable of the output layer, Represents the weight matrix of the output layer, Represents the hidden state, and the hidden state output is: In formula (4), Represents the weight matrix input to the hidden layer, Represents the weight matrix from the hidden layer to the hidden layer, Represents the bias vector of the hidden layer, Represents the first detection error value, and the loss function is calculated through the mean square error loss as: In formula (5), N represents the number of samples, Represents the true value of the i-th sample, Represents the predicted value of the deep learning model for the i-th sample, and the weight matrix is further updated through the optimization algorithm as: In formula (6), Represents the original point weight matrix, Represents the learning rate, Represents the total value in the operation path, Represents the prediction threshold, Represents the gradient of the loss function with respect to the weight matrix. The principle of this method lies in its unique network structure, which can consider the previous input information when processing the current input. This is achieved by introducing recurrent connections in the hidden layer, enabling the recurrent neural network to capture the temporal dependencies in sequential data. In the present invention, the power station operation data for the past year is collected for training the model, including the time series data of power output, temperature, humidity, light intensity, voltage, ripple, and current. The training parameter learning rate is configured to be 0.001, and the number of iterations is 100 times. During the training process, the recurrent neural network uses the backpropagation algorithm and the gradient descent optimization method to calculate the output and the new hidden state at each time step based on the input and the hidden state of the previous time step, and continuously adjusts the network parameters to minimize the prediction error. In this way, the recurrent neural network can learn how to extract useful features from the input data.

[0017] The hardware conditions for the above implementation are as follows: The supporting hardware architecture design of the deep learning model needs to combine computing efficiency, storage optimization, and low-power characteristics, and its core components include a variety of dedicated computing units and collaborative technologies. In terms of the core computing unit, the GPU (such as NVIDIA A100) has become the main force for training and inference in the data center due to its high parallel computing ability, supporting CUDA-accelerated LSTM matrix operations; the TPU (such as Google TPU v4) is optimized for cloud inference and achieves low power consumption and high throughput through tensor computing; the FPGA (such as the Xilinx Versal series) meets the low-latency requirements at the edge through a reconfigurable architecture and dynamically adapts to the RNN structure; the ASIC (such as Groq TSP) and NPU (such as HiSilicon Ascend 310) further improve the energy efficiency ratio and are optimized for specific scenarios and embedded devices respectively. In terms of memory and storage, high-bandwidth memory (HBM3) is paired with GPU / TPU to reduce off-chip access latency, and on-chip caches (such as Xilinx UltraRAM) optimize the access to timing data; low-power design uses dynamic voltage and frequency scaling (DVFS) and approximate computing (such as INT8 quantization) to reduce energy consumption. Special accelerators such as the neuromorphic architecture of Intel Loihi 2 simulate the pulse computing of deep learning algorithms to improve the timing processing efficiency. In terms of the interconnection architecture, NVLink and CXL technologies support multi-chip expansion, and near-memory computing (such as GDDR6-AiM) reduces data movement. In typical applications, cloud clusters (such as DGX A100) use the InfiniBand network to achieve distributed training, and edge devices (such as Zynq MPSoC) achieve microsecond-level inference through FPGA acceleration. Performance tests show that FPGA / ASIC is superior to GPU in terms of latency and energy efficiency ratio in edge scenarios, while GPU is more suitable for long-sequence batch processing, and TPU performs outstandingly in cloud inference. In step four, the quantum coding module uses quantum ground state coding to convert the distributed photovoltaic power station fault data information into a quantum state, the quantum measurement module uses a controlled-swap gate to calculate the similarity between the data points and the cluster centers, and the cluster data feature calculation module uses a quantum minimum-finding algorithm to find the most similar cluster center point, realizing the efficient identification of different types of distributed photovoltaic power station fault data information.

[0018] In a specific embodiment, the quantum encoding module is a bridge between quantum computing and classical computing. It completes the superposition and entanglement of quantum states through quantum gate operations, enabling quantum algorithms to process actual data sets. In the fault analysis of distributed photovoltaic power stations, a large number of data points need to be processed. The use of quantum ground state encoding reduces the use of qubits through a simple mapping relationship, lowering the requirements of the algorithm for quantum hardware. The quantum ground state encoding method is more stable and easier to implement and control. The quantum measurement module encodes data points and cluster centers into multiple quantum registers, and uses an additional control qubit to trigger a controlled swap gate. When the control qubit is in a specific state, the controlled swap gate will swap the states of the data point and cluster center encoding qubits. When the control qubit is in another state, no swapping occurs. According to the obtained similarity, the quantum minimum finding algorithm is used to find the most similar cluster center point. Through the parallelism and superposition of quantum computing, the minimum distance from each data point to each cluster center can be found in a shorter time, which can reduce the computational complexity and meet the large number of features and complex structures existing in the faults of distributed photovoltaic power stations.

[0019] The above-mentioned schemes are compared and analyzed. The traditional K-Means (implemented by Scikit-learn), quantum K-Means (Qiskit simulator, 4 qubits), and quantum heuristic algorithm (TensorFlow Quantum hybrid model) are compared and analyzed. Evaluation metrics: clustering accuracy rate (ARI index), number of convergence iterations, and single-iteration time. The results are shown in Table 1. Through experiments, the clustering accuracy rate of quantum K-Means is improved by 23.6% because quantum parallelism is better at capturing high-dimensional non-linear features; the convergence speed is accelerated by 46.7% because quantum measurement directly extracts the tendency of global optimality. A coding method based on dynamic rotation angle adjustment is proposed to adapt to the non-uniform distribution characteristics of photovoltaic data, and the coding efficiency is improved by 40%.

[0020] In step four, the method for the quantum K-Means algorithm model to realize the diagnosis and analysis of the fault data information of the distributed photovoltaic power station is: input data set n is the scale of the data set, the data set is divided into k categories, and the cluster center is Then the quantum states of the data point and the cluster center are: In formula (7), represents the j-th eigenvalue of the i-th data point, d represents the number of eigenvalues. In formula (8), represents the j-th eigenvalue of the i-th data point, k represents the number of dataset categories, and the similarity between any data point and k clustering centers is calculated by the controlled swap gate Control-Swap and k clustering centers: In formula (9), represents the cosine value of with c, and using the quantum state as the input of the phase estimation algorithm, we can obtain: In formula (10), the similarity between is stored in the quantum bit on, the smaller, the greater the similarity between the two. The steps to find the minimum value from the quantum state using the quantum minimum finding algorithm are as follows: (1) Prepare the quantum state of the initial value as (2) as the input, b as the control input, and use the Grover algorithm to find (3) If then use and go back to step (1), repeat times. The quantum K-Means algorithm model is a model that combines quantum computing and the classical K-Means clustering algorithm. Quantum computing is a new computing mode that calculates according to the quantum mechanics to control the quantum information unit quantum bit, which can break through the classical computing power bottleneck and has more powerful computing capabilities. The K-Means algorithm, also known as K-average or K-means, is a classical clustering algorithm. Its core idea is to divide the data into k independent clusters, making the distance between data points within each cluster as small as possible, and the distance between clusters as large as possible. By encoding data points and clustering centers into quantum states through quantum computing, using the controlled swap gate to calculate the similarity between data points and clustering centers, the process of phase estimation is mainly the process of Grover iteration, which requires corresponding Oracle operations, where the preparation of It requires 6 Oracle operations to find the minimum value from the quantum state through the quantum minimum search algorithm. The corresponding similarity between the two is the largest. Thus, the corresponding data points are classified into the corresponding cluster centers to achieve fault classification. Quantum computing is used to implement cluster update to accelerate the clustering process. The clustering method is the maximum-minimum distance clustering algorithm, which classifies the fault information elements of distributed photovoltaic power stations of various types. The core idea of this method is to first calculate the cluster centers, and then assign all sample points to the class corresponding to the nearest cluster center according to the principle of proximity. Maximum-minimum means selecting the largest among all the minimum distances. The main algorithm steps are as follows: (1) Randomly select a point as the cluster center Z1 of the first class; the multi-dimensional features of photovoltaic fault data (such as voltage, current, temperature, irradiance, etc.) are encoded into quantum states through parametric angles.

[0021] (2)Select the sample point farthest from the one in step 1 as the cluster center of the second class Prepare the quantum bits into superposition states through Hadamard gates (H gates), for example , to achieve parallel processing of all feature dimensions.

[0022] (3)Calculate the distance from each point to all cluster centers one by one, and record all the shortest distances; (4)Select the largest value among these shortest distances. If this maximum value is greater than where then take the other sample point corresponding to this maximum distance as the new cluster center, otherwise end the entire algorithm; (5) Repeat the operations in steps (3) and (4) until no new cluster center appears in (4); (6) Assign all samples to the cluster center closest to themselves. The above scheme is applied to the inverter operation data of a certain distributed photovoltaic power station, which includes 10 phalanxes and 30 inverters, with a collection period of 10 minutes, lasting for 1 month, and a total of 8,994,194 records. Feature extraction is performed on the data information, such as active power, DC voltage, AC current, irradiance, module temperature, ambient temperature, power volatility, temperature gradient, etc. Normalize each feature dimension, set parameter thresholds (such as the power threshold is ±20% of the rated power), use the Qiskit framework to construct a quantum circuit, and encode each data point into 3 qubits (corresponding to 3 feature dimensions). Randomly select 3 quantum states as the initial cluster centers (K = 3), perform the C-SWAP operation on each data point, and calculate the similarity with the 3 centroids. The cluster center with the highest measurement result probability is selected to complete the data point assignment. The hardware platform is the IBM Quantum Experience simulator (5 qubits), and the comparison algorithms are: classical K-Means (implemented by Python Scikit-learn) and quantum K-Means based on parametric angle encoding (Qiskit simulation). The experimental results are shown in Figure 2. It contains 32% of the data points, with the characteristics of power volatility > 15% and temperature gradient > 5°C / min, corresponding to the overheating fault of the inverter IGBT module. Cluster 2 (low irradiance anomaly): It accounts for 45% of the data points, with the characteristics of irradiance < 50 W / m² but power output > 10% of the rated value, corresponding to the sensor offset fault. Cluster 3 (normal operation): The remaining 23% of the data points, with the characteristic distribution within the normal range. Through quantum advantage analysis, the quantum measurement module simultaneously processes the comparison of 3 centroids through the superposition state, while the classical algorithm needs to calculate the distance serially 3 times. The quantum entanglement property (such as the multi-body entanglement state) can improve the robustness of data point assignment and still maintain a high purity in the presence of sensor noise.

[0023] In a further embodiment, the quantum K-Means algorithm model is used for the fault analysis of the distributed photovoltaic power station. The data analysis time of the faults in the distributed photovoltaic power station is shown in Table 3: According to the different scales of the data in the data group, four test groups are set up, and two methods are used to process the data. Method 1 is to process the data using the classical K-Means algorithm model, and Method 2 is to process the data using the quantum K-Means algorithm model. As shown in Table 1, when processing the first group of data, there is no obvious difference in the time used by Method 1 and Method 2. When processing the second group of data, the time used by Method 2 is reduced by about 15% compared to Method 1. When processing the third group of data, the time used by Method 2 is reduced by about 19% compared to Method 1. When processing the fourth group of data, the time used by Method 2 is reduced by about 60% compared to Method 1. It can be seen that the larger the data volume, the shorter the data processing time used by the method of the present invention, and the better the stability of the device. Although the specific embodiments of the present invention have been described above, those skilled in the art should understand that these specific embodiments are only examples. Without departing from the principle and essence of the present invention, those skilled in the art can make various omissions, substitutions, and changes to the details of the above methods and systems. For example, combining the above method steps, so as to perform substantially the same function according to a substantially same method to achieve substantially the same result, then it belongs to the scope of the present invention. Therefore, the scope of the present invention is only defined by the appended claims.

Claims

1. A method for fault analysis of a distributed photovoltaic power station based on a quantum K-Means algorithm model, characterized in that: It includes the following steps: Step 1: Collect the fault data information of the distributed photovoltaic power station through the data acquisition module. The data information includes at least power output, temperature, humidity, light intensity, voltage, ripple, and current. The data acquisition module includes a sensor module, a data preprocessing module, and a data conversion module to achieve real-time collection of data information; Step 2: Construct a blockchain network model. Transmit the fault data information of the distributed photovoltaic power station through the blockchain network model, package the processed data into transactions for data uploading to the chain, and achieve the security of data information. The blockchain network model includes a blockchain layer, a consensus mechanism, a data layer, and a network layer; Step 3: Extract the fault data information of the distributed photovoltaic power station through a feature extraction model, and automatically learn the deep feature representation of the data through a deep learning model to extract higher-level abstract features from the original data; Step 4: Diagnose and analyze the fault data information of the distributed photovoltaic power station through a quantum K-Means algorithm clustering model. Through the state superposition and quantum entanglement characteristics of quantum bits, achieve efficient search and clustering of multi-dimensional data. The quantum K-Means algorithm clustering model includes a quantum encoding module, a quantum measurement module, and a clustering data feature calculation module. The output end of the quantum encoding module is linked to the input end of the quantum measurement module, and the output end of the quantum measurement module is linked to the input end of the clustering data feature calculation module; Step 5: Establish a fault prediction model based on the clustering results. By receiving the real-time data information of the distributed photovoltaic power station, analyze and predict the future faults of the power station.

2. The distributed photovoltaic power station fault analysis method of a quantum K-Means algorithm model according to claim 1, characterized in that: The sensor module includes power output, temperature, humidity, light intensity, voltage, ripple, and current sensors. The data preprocessing module eliminates noise and interference through a first-order low-pass filter, and the data conversion module performs data conversion through the ADC analog-to-digital conversion formula.

3. A distributed photovoltaic power station fault analysis method based on the quantum K-Means algorithm model according to claim 1, characterized in that: The construction of the blockchain network model includes the following steps: Step 1: Select the Hyperledger blockchain platform and perform network configuration according to the actual needs of the distributed photovoltaic power station; Step 2: Set multiple nodes in the photovoltaic power station, including data collection nodes, analysis nodes, and verification nodes; Step 3: Select a proof-of-work consensus mechanism according to the network scale and performance requirements to ensure the security and reliability of the data in the network.

4. A distributed photovoltaic power station fault analysis method based on the quantum K-Means algorithm model according to claim 1, characterized in that: The quantum encoding module uses quantum ground state encoding to convert the fault data information of the distributed photovoltaic power station into a quantum state. The quantum measurement module uses a controlled swap gate to calculate the similarity between the data points and the clustering center. The clustering data feature calculation module uses a quantum minimum search algorithm to find the most similar clustering center point to achieve efficient identification of different types of fault data information of the distributed photovoltaic power station.

5. The distributed photovoltaic power station fault analysis method of a quantum K-Means algorithm model according to claim 1, characterized in that: The data acquisition module realizes the acquisition of data information including the following steps: Step 1: Eliminate noise and interference through a first-order low-pass filter. The recurrence formula in the discrete time domain can be expressed as: In formula (1), x[n] is the input signal, y[n] is the filtered output signal, and α is the coefficient of the filter, which determines the cut-off frequency and attenuation characteristics of the filter; Step 2: Implement data information conversion through the ADC analog-to-digital conversion formula, convert the analog voltage into a digital code, and the output information content is represented by the formula: In formula (2), where Vmin and Vmax are the input voltage ranges of the ADC, and N is the number of bits of the ADC. After passing through formula (2), the conversion of the analog voltage into a digital code can be achieved, and further, the processing and calculation of analog signals can be realized.

6. The distributed photovoltaic power station fault analysis method of a quantum K-Means algorithm model according to claim 1, characterized in that: The deep learning model in the feature extraction model is a recurrent neural network deep learning model.

7. A distributed photovoltaic power station fault analysis method based on the quantum K-Means algorithm model according to claim 5, characterized in that: The method for the recurrent neural network deep learning model to complete feature extraction is as follows: The basic structure of the recurrent neural network includes an input layer, a hidden layer, and an output layer. At each time step (t), the RNN receives an input , and updates its hidden state , and then generates an output . Then, the data information recognition function of the device in the industrial Internet environment after learning is: In formula (3), where represents the weight matrix from the hidden layer to the output layer, represents the bias variable of the output layer, represents the weight matrix of the output layer, represents the hidden state, and the hidden state output is: In formula (4), represents the weight matrix input to the hidden layer, represents the weight matrix from the hidden layer to the hidden layer, represents the bias vector of the hidden layer, represents the first detection error value, and the loss function is calculated by the mean squared error loss as: In formula (5), N represents the number of samples, represents the true value of the i-th sample, represents the predicted value of the i-th sample by the deep learning model, and further updates the weight matrix through an optimization algorithm as: In formula (6), represents the original point weight matrix, represents the learning rate, represents the total value in the operation path, represents the prediction threshold, represents the gradient of the loss function with respect to the weight matrix.

8. The distributed photovoltaic power station fault analysis method of a quantum K-Means algorithm model according to claim 1, characterized in that: The method for the quantum K-Means algorithm model to realize the diagnosis and analysis of the fault data information of the distributed photovoltaic power station is as follows: input the data set n is the scale of the data set, the data set is divided into k categories, and the cluster centers are Then the quantum states of the data points and the cluster centers are: In formula (7), represents the j-th eigenvalue of the i-th data point, d represents the number of eigenvalues. In formula (8), represents the j-th eigenvalue of the i-th data point, k represents the number of dataset categories. The similarity between any data point and k clustering centers is calculated by the controlled swap gate Control-Swap: In formula (9), express and the cosine value of c, in quantum state As input to the phase estimation algorithm, we get: In formula (10), The similarity between is stored in the qubit The smaller is, the greater the similarity between the two. The steps to find the minimum value from the quantum state using the quantum minimum finding algorithm are as follows: (1) Prepare the quantum state of the initial value as Take as the input, b as the control input, and use the Grover algorithm to find If then use and go back to step (1), repeating

Images