Hidden Markov-based distributed energy operation state prediction method and system

Abstract

The invention discloses a distributed energy operation state prediction method and system based on hidden Markov, and relates to the technical field of new energy and power system management, and the method comprises the steps: collecting the historical operation data of a distributed photovoltaic power generation system; adopting a GMM model to extract state feature data from the historical operation data, and performing operation state classification; optimizing the hidden Markov model by adopting a GA (Genetic Algorithm), and inputting the classified state feature data into the optimized GA-HMM for training; and inferring the current state of the system by utilizing the observation probability matrix and the state transition probability matrix of the HMM model and combining real-time data. According to the method, the GMM model is combined to classify the running states of the distributed photovoltaic power generation system and extract the state features, the HMM model is optimized through the GA algorithm, accurate recognition of the three running states of the distributed photovoltaic power generation system is achieved, and 100% accurate recognition is achieved in the aspect of fault state recognition.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy and power system management, and in particular to a method and system for predicting the operating state of distributed energy based on hidden Markov. Background Art

[0002] With the widespread application and rapid development of distributed energy systems, the demand for accurate prediction of system operating status is becoming increasingly urgent. However, due to the complexity and uncertainty of distributed energy systems, traditional prediction methods often fail to achieve ideal prediction results. Therefore, it is necessary to explore new prediction methods and models to improve prediction accuracy and reliability. As an effective time series data modeling and prediction tool, the hidden Markov model has unique advantages in predicting the operating status of distributed energy systems. HMM can capture the potential unobserved variables and state transition laws in the system by analyzing and modeling the historical operating data of the system, thereby accurately predicting the future state of the system. Summary of the invention

[0003] In view of the above-mentioned problems, the present invention is proposed.

[0004] Therefore, the technical problem solved by the present invention is: how to accurately predict the operating status of a distributed photovoltaic power generation system, combine historical data with real-time data, and use GMM-HMM model optimization to avoid local optimal problems, thereby achieving dynamic monitoring of the system status and fault prevention, and ensuring efficient and safe operation of the system.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a distributed energy operation state prediction method based on hidden Markov, which comprises the following steps:

[0006] Collect historical operation data of distributed photovoltaic power generation systems;

[0007] The GMM model is used to extract status feature data from historical operation data and classify the operation status;

[0008] The GA algorithm is used to optimize the hidden Markov model, and the classified state feature data is input into the optimized GA-HMM model for training to obtain a GA-HMM model with output of the operating state category of the distributed photovoltaic power generation system;

[0009] The observation probability matrix and state transition probability matrix of the HMM model are used in combination with real-time data to infer the current state of the system.

[0010] As a preferred solution of the distributed energy operation status prediction method based on hidden Markov described in the present invention, the historical operation data includes overall power output, current and voltage data, temperature data, equipment status data, weather and environmental data, power generation and revenue data, alarm and fault data.

[0011] As a preferred solution of the distributed energy operation state prediction method based on hidden Markov in the present invention, wherein: the operation state classification is to combine the equipment failure state of the historical operation data of the distributed photovoltaic power generation system, use the Gaussian mixture model to quantify the equipment failure, each Gaussian density function represents the characteristics of the operation state, estimate the mean of each state, and use the variance as the extraction of the state of the distributed photovoltaic power generation system;

[0012] The probability density function of the Gaussian mixture model is expressed as:

[0013]

[0014] Among them, α i is the mixing weight of the ith distribution, μ is the d-dimensional vector representing the mean of the distribution, and ∑ is the covariance matrix;

[0015] Assume K clusters, where a single Gaussian function is:

[0016]

[0017] Let α1α2α3 represent the probability that a number randomly selected initially belongs to the three states of normal, detection required and fault, and use the EM algorithm to calculate the parameter value of GMM;

[0018] The Gaussian mixture distribution parameters are known from the training data set, and the posterior probability of the i-th multivariate Gaussian distribution of each sample is obtained according to Bayes' theorem, expressed as:

[0019]

[0020] Among them, α j is the mixing weight of the jth distribution, μ i is the mean, x is a given data point, and is the posterior probability of the point drawn from one of the K Gaussian distributions.

[0021] Update the mean and variance, the formula is as follows:

[0022]

[0023]

[0024] Among them, γ is the posterior probability, μ' iis the updated mean, α' i is the mixed weight of the i-th distribution after update, i, j are, respectively, x j is the given data point of the i-th multivariate Gaussian distribution, m is the total number of samples, and is the total number of samples of the i-th Gaussian distribution.

[0025] As a preferred solution of the distributed energy operation state prediction method based on hidden Markov in the present invention, wherein: the GA-HMM model with output distributed photovoltaic power generation system operation state category includes optimizing the hidden Markov model by using GA algorithm;

[0026] Aiming at the situation that the initial value of the HMM model falls into the local optimum during parameter optimization, the GA algorithm is optimized, and the optimized model is the GA-HMM model.

[0027] As a preferred solution of the distributed energy operation state prediction method based on hidden Markov in the present invention, wherein: the GA-HMM model with output distributed photovoltaic power generation system operation state category also includes:

[0028] Genetic algorithm is used to encode the optimization object. The solution process uses the coding representation of the solution to encode the probability π, the initial state shift matrix A of the state transition, and the observation probability matrix B in the HMM model, and ensure that the sum of each row element of π, A, and B is 1. The constraints are:

[0029]

[0030] Among them, π i The probability of transitioning from the initial state to the i-th state, π is a vector of length N, α ij The probability of transitioning from state i to state j, b ij are the probabilities of observing state j in state i;

[0031] Construct a fitness function. The effect of the hidden Markov model is inversely proportional to the fitness function. In the Baum-Welch algorithm, find the parameters that maximize the conditional probability P = (O|λ0. The fitness function takes f(λ) = log(P(O k )|λ);

[0032] The preset maximum evolutionary generation is used as the termination condition, and GA optimization is performed on the parameter B in the HMM. As a preferred solution of the distributed energy operation status prediction method based on hidden Markov in the present invention, wherein: the GA optimization of the parameter B in the HMM includes determining the number of HMMs and the number of states of each HMM model, and converting them into corresponding HMM chromosomes for encoding, creating an initial population, and determining the fitness function f(λ)=log(P(O k)|λ), and calculate the fitness of the individual to determine whether the preset maximum number of iterations has been reached. If it has been reached, the optimization process ends. Otherwise, the next step is to perform selection, crossover, mutation and other operations on the population to form a new population. The fitness function is calculated for the new population and the maximum number of iterations is determined again.

[0033] When optimizing the B parameter in HMM model parameter learning, the GA genetic algorithm is introduced to construct a fitness function with the HMM model. The optimal solution is obtained by seeking the maximum value of the fitness function to form an optimized GA-HMM model.

[0034] As a preferred solution of the distributed energy operation state prediction method based on hidden Markov in the present invention, the observation probability matrix and state transition probability matrix of the HMM model are used to infer the current state of the system in combination with real-time data, including:

[0035] The real-time monitored electrical parameter values ​​are used as observation sequences to calculate the state transfer matrices. When the distributed photovoltaic power generation system is in a fault state, if the external manual maintenance is performed and the state cannot be transferred to the state requiring detection or the normal state, the state transfer probability calculation result is 0.

[0036] The possible state of the system in the future is predicted based on the current state and the state transition probability matrix, and the actual needs are met by setting different prediction time windows and prediction accuracies.

[0037] Another object of the present invention is to provide a distributed energy operation status prediction system based on hidden Markov, which can effectively realize the classification of equipment status and real-time prediction of system status by combining historical operation data with real-time data, thereby overcoming the problems of insufficient accuracy, poor real-time performance and unsatisfactory optimization effect in existing methods.

[0038] In order to solve the above technical problems, the present invention provides the following technical solutions: A distributed energy operation status prediction system based on hidden Markov includes: a data collection and preprocessing module, a state feature extraction and classification module, a GA-HMM model training and optimization module, and a real-time state prediction and decision support module;

[0039] The data collection and preprocessing module collects and organizes historical operation data, and cleans and normalizes the original data;

[0040] The state feature extraction and classification module uses a Gaussian mixture model to extract features and cluster the states in the operating data. Each operating state is quantified by the mean and variance of the Gaussian distribution, and the EM algorithm is used to optimize the parameters of the GMM model.

[0041] The GA-HMM model training and optimization module uses a genetic algorithm to optimize the traditional hidden Markov model, encodes and optimizes the initial state matrix, state transition matrix and observation probability matrix of the HMM, creates a fitness function, optimizes the evolutionary algebra of the GA, and makes the model parameters reach the optimal solution;

[0042] The real-time state prediction and decision support module infers the state of the current system based on real-time monitoring data combined with the state transition probability matrix of HMM.

[0043] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the distributed energy operation state prediction method based on hidden Markov are implemented as described above.

[0044] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the distributed energy operation status prediction method based on hidden Markov.

[0045] The beneficial effects of the present invention are as follows: by combining the GMM model to classify the operating status of the distributed photovoltaic power generation system and extracting the status features, and optimizing the HMM model through the GA algorithm, the three operating statuses of the distributed photovoltaic power generation system can be accurately identified, and 100% accurate identification can be achieved in the identification of the fault status. The present invention can identify the operating status of the system in real time, issue an early warning before the system enters the fault state, reduce downtime and losses, and provide data support and decision-making basis for equipment maintenance, system upgrades and long-term planning. The distributed energy system operating status prediction method based on the hidden Markov model can not only realize the real-time monitoring and fault warning of the system, but also provide strong support for the system's optimized scheduling and long-term planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0047] Figure 1 A HMM model parameter flattening coding diagram of a distributed energy operation state prediction method based on hidden Markov provided in the first embodiment of the present invention;

[0048] Figure 2 A flow chart of a GA optimized HMM model in a distributed energy operation state prediction method based on hidden Markov provided in the first embodiment of the present invention;

[0049] Figure 3 A state recognition accuracy diagram of 36 sets of data of the distributed energy operation state prediction method based on hidden Markov provided in the first embodiment of the present invention. DETAILED DESCRIPTION

[0050] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0051] Example 1, reference Figure 1 to Figure 3 As an embodiment of the present invention, a distributed energy operation state prediction method based on hidden Markov is provided, comprising:

[0052] S01. Collect historical operation data of distributed photovoltaic power generation systems.

[0053] Overall power output: monitors the overall power output of the photovoltaic power generation system to evaluate the power generation capacity and efficiency of the system. Inverter power: includes the input power and output power of the inverter. These data help to understand the conversion efficiency and performance of the inverter.

[0054] Current and voltage data. DC side voltage and current: monitor the DC voltage and current generated by the photovoltaic cell modules to ensure the normal operation of the system.

[0055] AC side voltage and current: Monitor the AC voltage and current output by the inverter, as well as the voltage and current of the grid to ensure the safe transmission and distribution of electrical energy.

[0056] Temperature data. Photovoltaic cell module temperature: monitor the temperature changes of photovoltaic cell modules to promptly detect problems of excessively high or low temperatures to avoid affecting power generation efficiency and equipment life.

[0057] Ambient temperature: Monitor the ambient temperature around the photovoltaic power station to provide a reference for system performance evaluation and operation and maintenance.

[0058] Equipment status data. Photovoltaic cell module status: including the module's power generation efficiency, health status, etc.

[0059] Inverter status: monitor the operating status and performance indicators of the inverter, such as conversion efficiency, fault alarm, etc.

[0060] Energy storage system status: For photovoltaic power generation systems equipped with energy storage systems, it is also necessary to monitor the charging and discharging status and remaining power of the energy storage batteries.

[0061] Weather and environmental data.

[0062] Light intensity: Monitors the light intensity of a PV plant to assess the system’s power generation potential and performance.

[0063] Wind speed and direction: For large-scale ground-based photovoltaic power stations, wind speed and direction data are of great significance for evaluating the impact of wind pressure on photovoltaic brackets and optimizing the layout of power stations.

[0064] Humidity and air pressure: These data may affect the performance and life of photovoltaic equipment, so they also need to be monitored.

[0065] Power generation and revenue data.

[0066] Power generation statistics: including daily, monthly, annual power generation, and cumulative power generation, etc., used to evaluate the power generation capacity and economic benefits of the system.

[0067] Power generation revenue: Calculate the power generation revenue of the photovoltaic power generation system based on electricity price policy and market conditions.

[0068] Alarm and fault data.

[0069] Real-time alarm information: When the monitoring system detects an abnormal situation or failure, it will promptly send an alarm message to the operation and maintenance personnel so that the problem can be dealt with in time to avoid losses.

[0070] Fault recording and diagnosis: Record the time, cause and handling of system faults to provide reference for subsequent operation, maintenance and optimization.

[0071] The operating status data of distributed photovoltaic power generation systems covers power, current and voltage, temperature, equipment status, weather and environment, power generation and revenue, alarms and faults, etc. These data are collected in real time through hardware devices such as data collectors and sensors.

[0072] S02. Use the GMM model to extract status feature data from historical operation data and classify the operation status;

[0073] Combined with the equipment failure status of the historical operation data of the distributed photovoltaic power generation system, the Gaussian mixture model can be used to quantify the equipment failure. The Gaussian mixture model decomposes the overall data into the additive form of multiple Gaussian probability density functions. In the present invention, the operating status of the distributed photovoltaic power generation system is "normal", "need to be detected" and "faulty". Each Gaussian density function represents the characteristics of the operating state, so as to estimate the mean of each state, and use the variance as the extraction of the state of the distributed photovoltaic power generation system.

[0074] Regarding the Gaussian mixture model (GMM), its probability density function is as follows:

[0075]

[0076] α i is the mixing weight of the ith distribution, summing to 1. Where μ is a d-dimensional vector representing the mean of the distribution, and ∑ is the covariance matrix. Assume there are K clusters. The single Gaussian function is:

[0077]

[0078] Let α1α2α3 represent the probability that a number randomly selected at the beginning belongs to the three states of "normal", "need to be detected" and "fault". We use the EM algorithm to calculate the parameter values ​​of GMM. The EM algorithm is an iterative method that can find the maximum likelihood estimate of the model parameters when there are some hidden parameters.

[0079] Assuming that the parameters of the Gaussian mixture distribution are known based on the training data set, the posterior probability of the i-th multivariate Gaussian distribution of each sample can be obtained according to Bayes' theorem:

[0080]

[0081] In this way, the mean and variance are updated. The formula is as follows:

[0082]

[0083] Gaussian distribution can accurately represent the parameters of distributed photovoltaic power generation systems under various operating conditions and can play a role in classifying operating conditions. However, the Gaussian mixture model can only extract the state characteristics of distributed photovoltaic power generation systems in transformers, but cannot identify the meaning of each Gaussian cluster. A method that can identify the various operating conditions of distributed photovoltaic power generation systems and achieve the function of a classifier is needed.

[0084] After evaluating its own model parameters, the optimized Baum-Welch algorithm in the GA-HMM model uses the forward-backward algorithm to calculate the probability of different operating states (normal, inspection required, faulty) under a given set of observation sequences, and infers the operating state of the distributed photovoltaic power generation system by comparing the results. Based on this principle, the GA-HMM model learns the degradation mode of the operating state of the distributed photovoltaic power generation system based on the operating data of the distributed photovoltaic power generation system, obtains the parameter λ of different operating states, and provides a degraded model library for the test phase. The real-time observation data is input into each model, the most likely state under the observation sequence is calculated, and the operating state classification is completed.

[0085] Cluster analysis is performed on the case data to obtain the recognition pattern of the operating status of the distributed photovoltaic power generation system. Its concentration distribution is obtained after clustering with the Gaussian mixture model. Different colors are set to indicate the color concentration distribution diagram when the distributed photovoltaic power generation system is in normal, requires inspection, or faulty operating status.

[0086] S03, using GA algorithm to optimize the hidden Markov model (HMM model), inputting the classified state feature data into the optimized GA-HMM model for training, and obtaining a GA-HMM model with an output distributed photovoltaic power generation system operation state category;

[0087] The GA algorithm is used to optimize the hidden Markov model (HMM model). The flowchart of GA optimization of the HMM model is as follows Figure 2 Specifically:

[0088] In view of the situation that the initial value of the HMM model falls into the local optimum when searching for parameters, the GA algorithm is used to optimize it. The optimized model is called GA-HMM. It should be noted that it should be adapted to the actual situation of the HMM model. The following introduces the selection method of important parameters:

[0089] (1) Selection of chromosome code

[0090] like Figure 1 As shown in the figure. The genetic algorithm requires that the object to be optimized be encoded, and the solution process is represented by the encoding of the solution. In this paper, the probability π in the HMM model, the initial state shift matrix A of the state transition, and the observation probability matrix B are encoded. And it is ensured that the sum of each row element of π, A, and B is 1.

[0091] The constraints are:

[0092]

[0093] (2) Fitness function

[0094] The quality of the model depends on the value of the fitness function. The smaller the value of the fitness function, the better the model. Considering that the model is HMM, in the Baum-Welch algorithm, the goal is to find the parameters that maximize the conditional probability P = (O|λ), and the fitness function is f(λ) = log(P(O k )|λ), thus ensuring the optimal result.

[0095] (3) Genetic Operator

[0096] The present invention adopts the roulette method to select the operator, the mutation probability Pm of the mutation operator is taken as 0.005 under the conditions of this paper; and the crossover operator adopts arithmetic crossover.

[0097] (4) Termination conditions

[0098] The preset maximum evolutionary generation is used as the termination condition, and the maximum evolutionary generation G is set to 50 in this paper.

[0099] To optimize the parameter B in HMM using GA, the key steps are as follows:

[0100] (1) Determine the number of HMMs and the number of states of each HMM model, convert them into corresponding HMM chromosomes, encode them according to the above encoding method, and create an initial population;

[0101] (2) Determine the fitness function f(λ) = log(P(O k )|λ), and calculate the fitness of the individual;

[0102] (3) Check whether the preset maximum number of iterations has been reached. If not, proceed to the next step. If the maximum number of iterations has been reached, the optimization search ends.

[0103] (4) Perform operations such as selection, crossover, and mutation on the population to form a new population;

[0104] (5) Calculate the fitness function for the new population and jump to the third step.

[0105] By studying the three types of problems that need to be solved when solving the hidden Markov model parameters: evaluation, decoding, and model parameters, the GA genetic algorithm is introduced to address the fact that the B parameter in the HMM model parameter learning is sensitive and easy to fall into the local optimum when seeking optimization. After introducing the optimization process, a suitable fitness function is constructed with the HMM model. The optimal solution is obtained by seeking the maximum value of the fitness function, forming a GA-HMM model with a higher degree of optimization.

[0106] It should be further explained that the construction of a suitable fitness function can be to set the goal to maximize the log-likelihood of the HMM model or minimize the prediction error, and to construct the corresponding fitness function through this goal: given the B parameter, calculate the log-likelihood of the HMM model under the current parameters. The log-likelihood is processed to ensure that the fitness function value is within a reasonable range. The normalized log-likelihood is returned as the fitness value.

[0107] S04. Use the observation probability matrix and state transition probability matrix of the HMM model and combine them with real-time data to infer the current state of the system.

[0108] The real-time monitored electrical parameter values ​​are used as observation sequences to calculate the state transfer matrices. When the distributed photovoltaic power generation system is in a fault state, without external manual maintenance, its state cannot be transferred to the state that needs to be detected or

[0109] Normal state, and the state transition probability calculation result is also 0.

[0110] For example, the probability of a distributed photovoltaic power generation system transferring from a normal state to a healthy state is 0.88, the probability of transferring to a sub-healthy state is 0.06, and the probability of transferring to a fault state is 0.05; the probability of transferring from a sub-healthy state to a sub-healthy state is 0.85, and the probability of transferring to a fault state is 0.2; its transfer matrix A is:

[0111]

[0112] The possible state of the system in the future is predicted based on the current state and the state transition probability matrix, and different prediction time windows and prediction accuracy requirements are set to meet actual needs.

[0113] 36 sets of operating parameters of photovoltaic power generation systems in a certain area were collected for testing. The experimental data proved that the GA-HMM model can be used in the distributed photovoltaic power generation system state recognition model. The recognition results are as follows: Figure 3 As shown in the figure, the recognition rate of normal state is 80%, the recognition rate of state to be detected is 81%, and the fault recognition rate is 100%. Therefore, the GA-HMM model has a good effect in identifying the state, and this method can be applied to the state recognition of the health management of distributed photovoltaic power generation systems.

[0114] Example 2, reference Figure 2 , which is an embodiment of the present invention, provides a system for distributed energy operation status prediction method based on hidden Markov, including: data collection and preprocessing module, state feature extraction and classification module, GA-HMM model training and optimization module, real-time state prediction and decision support module.

[0115] The data collection and preprocessing module collects and organizes historical operation data, and cleans and normalizes the original data;

[0116] The state feature extraction and classification module uses a Gaussian mixture model to extract and cluster the states in the operating data. Each operating state is quantified by the mean and variance of the Gaussian distribution, and the EM algorithm is used to optimize the parameters of the GMM model.

[0117] The GA-HMM model training and optimization module uses genetic algorithms to optimize the traditional hidden Markov model, encodes and optimizes the initial state matrix, state transition matrix and observation probability matrix of the HMM, creates a fitness function, optimizes the evolutionary algebra of GA, and makes the model parameters reach the optimal solution;

[0118] The real-time state prediction and decision support module infers the state of the current system based on real-time monitoring data combined with the state transition probability matrix of HMM.

[0119] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program code.

[0120] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0121] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0122] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0123] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A distributed energy operation status prediction method based on hidden Markov is characterized by: include: Collect historical operation data of distributed photovoltaic power generation systems; The GMM model is used to extract status feature data from historical operation data and classify the operation status; The GA algorithm is used to optimize the hidden Markov model, and the classified state feature data is input into the optimized GA-HMM model for training to obtain a GA-HMM model with output of the operating state category of the distributed photovoltaic power generation system; The observation probability matrix and state transition probability matrix of the HMM model are used in combination with real-time data to infer the current state of the system.

2. The method for predicting the operating status of distributed energy based on hidden Markov model according to claim 1, characterized in that: The historical operation data includes overall power output, current and voltage data, temperature data, equipment status data, weather and environmental data, power generation and revenue data, alarm and fault data.

3. The method for predicting the operating status of distributed energy based on hidden Markov as claimed in claim 2, characterized in that: The operation status classification is to combine the equipment failure status of the historical operation data of the distributed photovoltaic power generation system, use the Gaussian mixture model to quantify the equipment failure, each Gaussian density function represents the characteristics of the operation status, estimate the mean of each status, and use the variance as the extraction of the distributed photovoltaic power generation system status; The probability density function of the Gaussian mixture model is expressed as: Among them, α i is the mixing weight of the ith distribution, μ is the d-dimensional vector representing the mean of the distribution, and ∑ is the covariance matrix; Assume K clusters, where a single Gaussian function is: Let α1α2α3 represent the probability that a number randomly selected initially belongs to the three states of normal, detection required and fault, and use the EM algorithm to calculate the parameter value of GMM; The Gaussian mixture distribution parameters are known from the training data set, and the posterior probability of the i-th multivariate Gaussian distribution of each sample is obtained according to Bayes' theorem, expressed as: Among them, α j is the mixing weight of the jth distribution, μ i is the mean, x is a given data point, and is the posterior probability of the point drawn from one of the K Gaussian distributions. Update the mean and variance, the formula is as follows: Among them, γ is the posterior probability, μ' i is the updated mean, α' i is the mixed weight of the i-th distribution after update, i, j are, respectively, x j is the given data point of the i-th multivariate Gaussian distribution, m is the total number of samples, and is the total number of samples of the i-th Gaussian distribution.

4. The method for predicting the operating status of distributed energy based on hidden Markov as claimed in claim 3, characterized in that: The obtaining of the GA-HMM model with output of the operating status category of the distributed photovoltaic power generation system includes optimizing the hidden Markov model using a GA algorithm; Aiming at the situation that the initial value of the HMM model falls into the local optimum during parameter optimization, the GA algorithm is optimized, and the optimized model is the GA-HMM model.

5. The method for predicting the operating status of distributed energy based on hidden Markov as claimed in claim 4, characterized in that: The GA-HMM model with output of the operating status category of the distributed photovoltaic power generation system also includes: Genetic algorithm is used to encode the optimization object. The solution process uses the coding representation of the solution to encode the probability π, the initial state shift matrix A of the state, and the observation probability matrix B in the HMM model, and ensure that the sum of each row element of π, A, and B is 1. The constraints are: Among them, π i The probability of transitioning from the initial state to the i-th state, π is a vector of length N, α ij The probability of transitioning from state i to state j, b ij are the probabilities of observing state j in state i; Construct a fitness function. The effect of the hidden Markov model is inversely proportional to the fitness function. In the Baum-Welch algorithm, find the parameters that maximize the conditional probability P = (O|λ). The fitness function takes f(λ) = log(P(O k )|λ); The preset maximum evolutionary number is used as the termination condition, and GA optimization is performed on the parameter B in the HMM.

6. The method for predicting the operating status of distributed energy based on hidden Markov as claimed in claim 5, characterized in that: The GA optimization of the parameter B in the HMM includes determining the number of HMMs and the number of states of each HMM model, converting them into corresponding HMM chromosomes for encoding, creating an initial population, and determining the fitness function f(λ)=log(P( k )|λ), and calculate the fitness of the individual to determine whether the preset maximum number of iterations has been reached. If it has been reached, the optimization process ends. Otherwise, the next step is to perform selection, crossover, mutation and other operations on the population to form a new population. The fitness function is calculated for the new population and the maximum number of iterations is determined again. When optimizing the B parameter in HMM model parameter learning, the GA genetic algorithm is introduced to construct a fitness function with the HMM model. The optimal solution is obtained by seeking the maximum value of the fitness function to form an optimized GA-HMM model.

7. The method for predicting the operating status of distributed energy based on hidden Markov as claimed in claim 6, characterized in that: The observation probability matrix and state transition probability matrix of the HMM model are used to infer the current state of the system in combination with real-time data, including: The real-time monitored electrical parameter values ​​are used as observation sequences to calculate the state transfer matrices. When the distributed photovoltaic power generation system is in a fault state, if the external manual maintenance is performed and the state cannot be transferred to the state requiring detection or the normal state, the state transfer probability calculation result is 0. The possible state of the system in the future is predicted based on the current state and the state transition probability matrix, and the actual needs are met by setting different prediction time windows and prediction accuracies.

8. A system using the distributed energy operation state prediction method based on hidden Markov as claimed in any one of claims 1 to 7, characterized in that: It includes data collection and preprocessing module, state feature extraction and classification module, GA-HMM model training and optimization module, and real-time state prediction and decision support module; The data collection and preprocessing module collects and organizes historical operation data, and cleans and normalizes the original data; The state feature extraction and classification module uses a Gaussian mixture model to extract features and cluster the states in the operating data. Each operating state is quantified by the mean and variance of the Gaussian distribution, and the EM algorithm is used to optimize the parameters of the GMM model. The GA-HMM model training and optimization module uses a genetic algorithm to optimize the traditional hidden Markov model, encodes and optimizes the initial state matrix, state transition matrix and observation probability matrix of the HMM, creates a fitness function, optimizes the evolutionary algebra of the GA, and makes the model parameters reach the optimal solution; The real-time state prediction and decision support module infers the state of the current system based on real-time monitoring data combined with the state transition probability matrix of HMM.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the distributed energy operation status prediction method based on hidden Markov chain are implemented as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the distributed energy operation status prediction method based on hidden Markov in any one of claims 1 to 7 are implemented.

Images