Inverter operation state evaluation method, device and equipment based on data driving

Abstract

The invention discloses a method, a device and equipment for evaluating the running state of an inverter based on data driving. According to the method, the inverter operation state prediction model based on the multi-population collaborative CRPF and the inverter operation state error correction model based on the LSTM are utilized at the same time to predict the operation state of the inverter, and the inverter operation state prediction model based on the multi-population collaborative CRPF can predict the operation state of the inverter without a large amount of data. The LSTM-based inverter operation state error correction model has a gating mechanism and a long-term memory mechanism, can smoothly capture the mutual dependency relationship existing in the state data in the operation process of the inverter, and has good prediction accuracy through the combined use of the two models. The problem of low effectiveness of state evaluation caused by less historical data in the initial operation period of the inverter is solved, the relevance of the operation state data of the inverter is reserved, and the accuracy of the operation state evaluation of the inverter is further improved.

Description

Technical Field

[0001] The present invention relates to the field of electrical engineering technology, and in particular to a data-driven inverter operating status evaluation method, device and computer equipment. Background Art

[0002] As the new energy market prospers, inverter technology is also gradually innovating and improving, providing a strong guarantee for the efficient operation of new energy power generation and energy storage systems. As the core component of photovoltaic power generation and energy storage, the inverter converts DC photovoltaic power into AC power for input into the grid. Its stability, reliability and efficiency are crucial to the operation of the entire power generation and energy storage system.

[0003] However, in actual applications, photovoltaic inverters run for a long time in a complex working environment, and the inverters may have problems such as internal circuit failures and aging of various components, which will lead to the degradation of inverter performance and seriously affect the reliability and stability of photovoltaic inverter operation, thereby affecting the quality of the power grid or energy storage efficiency. Therefore, real-time monitoring and evaluation of the operating status of the inverter has important practical engineering significance.

[0004] At present, the inverter operation status assessment method based on deep learning is mainly used to monitor and evaluate the operation status of photovoltaic inverters. By training various neural network models, inverter operation status monitoring is realized. This type of method generally has high accuracy, but it relies on a large amount of real-time inverter operation data. During the initial operation of the inverter, the lack of historical data will cause low effectiveness of status assessment. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a data-driven inverter operating status evaluation method, device and computer equipment to solve the problem in the prior art that the historical data during the initial operation of the inverter is small and the status evaluation effectiveness is low.

[0006] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0007] In the first aspect, a data-driven inverter operation status assessment method is provided, including: obtaining status data of the inverter during operation; constructing an inverter operation status prediction model based on multi-population collaborative CRPF based on the status data, and constructing an inverter operation status error correction model based on LSTM; using the inverter operation status prediction model based on multi-population collaborative CRPF to calculate the state prediction value of the inverter within a specified time period; calculating the error between the state prediction value and the actual state value of the inverter within the specified time period; inputting the error into the inverter operation status error correction model based on LSTM to calculate the error prediction value; calculating the final state prediction value according to the state prediction value and the error prediction value, and evaluating the operation status of the inverter according to the final state prediction value.

[0008] In the second aspect, a data-driven inverter operation status evaluation device is provided, including: a data acquisition module, used to acquire the state data of the inverter during operation; a model construction module, used to construct an inverter operation status prediction model based on multi-population collaborative CRPF based on the state data, and to construct an inverter operation status error correction model based on LSTM; a state prediction module, used to calculate the state prediction value of the inverter within a specified time period using the inverter operation status prediction model based on multi-population collaborative CRPF; an error calculation module, used to calculate the error between the state prediction value and the actual state value of the inverter within the specified time period; an error prediction module, used to input the error into the inverter operation status error correction model based on LSTM to calculate the error prediction value; a state evaluation module, used to calculate the final state prediction value according to the state prediction value and the error prediction value, and evaluate the operation status of the inverter according to the final state prediction value.

[0009] In a third aspect, a computer device is provided, comprising at least one processor and at least one memory in communication with the processor, wherein the memory stores program instructions, and when the program instructions are called and executed by the processor, the above-mentioned data-driven inverter operating status assessment method is implemented.

[0010] In the scheme implemented by the above-mentioned data-driven inverter operation status evaluation method, device and computer equipment, by obtaining the status data of the inverter during operation, an inverter operation status prediction model based on multi-population collaborative CRPF is constructed based on the status data to predict the state prediction value of the inverter within a specified time period. At the same time, an inverter operation status error correction model based on LSTM is constructed based on the status data to predict the error between the state prediction value and the actual state value of the inverter within the specified time period to obtain an error prediction value. Finally, the final state prediction value is calculated based on the state prediction value and the error prediction value, and the operation status of the inverter is evaluated based on the final state prediction value, thereby realizing real-time monitoring and evaluation of the inverter operation status, timely discovering the abnormal operation status of the inverter, improving the stability of the inverter during operation, reducing maintenance costs, and improving power quality and related economic benefits. Compared with the existing inverter operation status assessment method, the present invention simultaneously utilizes two models, namely, the inverter operation status prediction model based on multi-population collaborative CRPF and the inverter operation status error correction model based on LSTM, to predict and deduce the operation status of the inverter, integrating the advantages of the two models. Among them, the inverter operation status prediction model based on multi-population collaborative CRPF can realize the prediction of the inverter operation status without a large amount of data, and has good prediction accuracy. The inverter operation status error correction model based on LSTM has a gating mechanism and a long-term memory mechanism, which can smoothly capture the interdependence of the state data during the operation of the inverter. The combined use of the two models not only overcomes the problem of low effectiveness of the state assessment caused by the small amount of historical data during the initial operation of the inverter, but also retains the correlation of the inverter operation status data, further improving the accuracy of the inverter operation status assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0012] Figure 1 is a flow chart of a method for evaluating inverter operation status based on data driving in one embodiment of the present invention;

[0013] Figure 2 This is the LSTM model architecture diagram;

[0014] Figure 3 This is a model loss function diagram of the inverter operation state error correction model based on LSTM;

[0015] Figure 4It is a diagram of the multi-population intelligent collaborative resampling mechanism;

[0016] Figure 5 is a structural schematic diagram of an inverter operation status evaluation device based on data drive in one embodiment of the present invention;

[0017] Figure 6 It is a structural diagram of a computer device in one embodiment of the present invention. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] The embodiments of the present application provide a data-driven inverter operating status evaluation method, device and computer equipment. By collecting the status data of the inverter during operation, the operating status of the inverter is deduced based on the status data, the stability of the inverter during operation is ensured, and the quality of the converted power grid is improved.

[0020] Figure 1 A flow chart of a method for evaluating inverter operation status based on data drive provided by an embodiment of the present invention is shown in FIG. Figure 1 As shown, the inverter operation status evaluation method based on data drive includes steps S10 to S60:

[0021] S10: Acquire status data of the inverter during operation.

[0022] The status data of the inverter may be output power data, or other status data such as output voltage data, output current data, etc. The output power data, output voltage data, and output current data of the inverter during operation may be collected by a power meter, a voltage meter, and an ammeter, respectively. A computer device (such as a server) for monitoring the operating status of the inverter may be connected to the power meter, the voltage meter, the ammeter, and other instruments in communication, so that the status data x of the inverter during operation may be obtained. T}, x k Represents the state data at time k, k=1,2,......,T.

[0023] S20: Construct an inverter operation status prediction model based on multi-population collaborative CRPF based on the status data, and construct an inverter operation status error correction model based on LSTM.

[0024] The process of constructing the inverter operation state prediction model based on multi-population coordinated CRPF is as follows: the state data x of the inverter during operation is obtained as follows: T The data is used as training data and input into a multi-population collaborative cost reference particle filter (CRPF) model for model training until the model converges to obtain a trained inverter operation status prediction model based on multi-population collaborative CRPF.

[0025] The process of constructing the inverter operation state error correction model based on LSTM is as follows: First, the state data of the inverter during operation x={x1,x2,......,x T} Input into the inverter operation state prediction model based on multi-population collaborative CRPF to obtain the state prediction value of the inverter during operation k=1,2,......,T; Then, calculate the error e between the predicted state value and the actual state value of the inverter during operation k , (k=1,2,...,T); finally, the error e k As training data, it is input into the long short-term memory networks (LSTM) model for model training until the model converges, and a trained LSTM-based inverter operation state error correction model is obtained.

[0026] As one of the many variants of recurrent neural networks, LSTM introduces gating mechanisms and memory units into its model, which can store and transmit important information in the data for a long time, effectively avoiding the gradient vanishing and gradient exploding problems that exist in traditional neural networks. At the same time, LSTM also has strong feature learning capabilities, which can automatically extract useful information contained in relevant data and further improve prediction accuracy. The LSTM model architecture is shown in the figure below. Figure 2 shown.

[0027] The specific training process of the inverter operation state error correction model based on LSTM is as follows: a single-input and single-output LSTM model is constructed, in which the number of hidden layer neuron nodes is 500, and the error e kAs training data, the number of training times is set to 2000, the initial learning rate is 0.1, and then the model training is started. After 200 rounds of training, in order to make the model gradually converge to obtain the optimal solution, the learning rate is reduced to 0.02. When performing model training and verification, a loss function is usually used to judge the effectiveness of the trained model. If the loss functions of the training set and the verification set converge to a certain effective value after several iterations, it means that the constructed LSTM-based inverter operating state error correction model can well characterize the state characteristics of the inverter during operation and perceive the changes in error data in time. When constructing the LSTM-based inverter operating state error correction model, the present invention selects the cross entropy group as the loss function (i.e. ) to train the model, such as Figure 3 As shown in the figure, it can be seen that after 800 rounds of training, the loss functions of the training set and the validation set converge to 0.5, and there is no significant change afterwards, indicating that there is a good fitting performance between the true error and the prediction error.

[0028] S30: Calculate the inverter operation state prediction value within a specified time period using the inverter operation state prediction model based on multi-population collaborative CRPF.

[0029] It is necessary to calculate the state prediction value of the inverter within a specified time period, and it is necessary to obtain the state data of the inverter during operation x={x1,x2,......,x T}, and input the inverter status data within the specified time into the inverter operation status prediction model based on multi-population collaborative CRPF, so as to obtain the state prediction value of the inverter within the specified time period. It should be noted that the inverter status data within the specified time period can be the obtained inverter status data during operation x={x1,x2,......,x T}, or may be the state data x={x1,x2,......,x T}.

[0030] S40: Calculate the error between the predicted state value and the actual state value of the inverter within a specified time period. The specific calculation method is as follows:

[0031]

[0032] Among them, e k represents the error between the predicted state value and the actual state value of the inverter at time k, represents the state prediction value of the inverter at time k, x k Represents the state data of the inverter at time k, which is the actual state value of the inverter at time k.

[0033] S50: Input the error into the inverter operation state error correction model based on LSTM to calculate the error prediction value.

[0034] Specifically, before calculating the error prediction value, first determine whether the error meets the error requirement. For example, if the error e at each moment is calculated k (k=1,2,......,T) are all less than 0.01, indicating that the error e k Meet the error requirements, otherwise, it means the error e k The error requirement is not met.

[0035] In the error e k When the error requirement is met, it indicates that the state prediction value of the inverter in the specified time period calculated by the inverter operation state prediction model based on multi-population collaborative CRPF in step 30 can characterize the operation state of the inverter during operation, and no error correction is required. Therefore, no error prediction value calculation is required; in the error e k If the error requirement is not met, it indicates that the state prediction value of the inverter in the specified time period calculated by the inverter operation state prediction model based on multi-population collaborative CRPF in step 30 cannot represent the operation state of the inverter during operation, and error correction is required. Therefore, error prediction value calculation is required: the error e calculated in step S40 is k Input the trained LSTM-based inverter operation state error correction model to calculate the error prediction value

[0036] It should be noted that, in other embodiments of the present invention, when executing step S50, it is not necessary to first determine whether the error meets the error requirement. Regardless of whether the error meets the error requirement, the error prediction value is calculated.

[0037] S60: Calculate a final state prediction value according to the state prediction value and the error prediction value, and evaluate the operating state of the inverter according to the final state prediction value.

[0038] Final state prediction value The calculation method is as follows:

[0039]

[0040] Among them, the state prediction value The error prediction value is calculated in step S30. It is calculated by step S50. It should be noted that, in the error e k When the error requirements are met, that is, when there is no need to calculate the error prediction value,

[0041] The specific method of evaluating the operating state of the inverter according to the final state prediction value is: preset a threshold range. If the final state prediction value falls within the threshold range, it indicates that the operating state of the inverter is normal; if the final state prediction value deviates from the threshold range, it indicates that the operating state of the inverter is abnormal.

[0042] The data-driven inverter operation status assessment method provided by the embodiment of the present invention obtains the state data of the inverter during operation, relies on the state data to construct an inverter operation status prediction model based on multi-population collaborative CRPF to predict the state prediction value of the inverter within a specified time period, and relies on the state data to construct an inverter operation status error correction model based on LSTM to predict the error between the state prediction value and the actual state value of the inverter within the specified time period to obtain an error prediction value, and finally calculates the final state prediction value according to the state prediction value and the error prediction value, and evaluates the operation status of the inverter according to the final state prediction value, thereby realizing real-time monitoring and evaluation of the inverter operation status, timely discovering the abnormal operation status of the inverter, improving the stability of the inverter during operation, reducing maintenance costs, and improving power quality and related economic benefits.

[0043] Compared with the existing inverter operation status assessment method, the data-driven inverter operation status assessment method provided by the embodiment of the present invention simultaneously utilizes two models, namely, the inverter operation status prediction model based on multi-population collaborative CRPF and the inverter operation status error correction model based on LSTM, to predict and deduce the operation status of the inverter, integrating the advantages of the two models. Among them, the inverter operation status prediction model based on multi-population collaborative CRPF can realize the prediction of the inverter operation status without a large amount of data, and has good prediction accuracy. The inverter operation status error correction model based on LSTM has a gating mechanism and a long-term memory mechanism, which can smoothly capture the interdependence of the state data during the operation of the inverter. The combined use of the two models not only overcomes the problem of low effectiveness of the state assessment caused by the small amount of historical data during the initial operation of the inverter, but also retains the correlation of the inverter operation status data, further improving the accuracy of the inverter operation status assessment.

[0044] In one embodiment, step S30 includes steps S31 to S35:

[0045] S31: Based on the state data, M populations are established, and the number of particles in each population is N.

[0046] Step S10 obtains the state data x of the inverter during operation = {x1, x2, ..., x T} k(k=1,2,......,T) obeys the uniform distribution U(I o ), will evenly distribute U(I o ) is evenly divided into M, and particles are extracted from the importance distribution of these M different initial values ​​to establish a population, and the number of particles in each population is N. Establish the particle-cost set of the M populations at the initial moment The particle cost of all populations at the initial moment Where i represents the particle number of the population, s represents the population number, i = 1, 2, ..., N, s = 1, 2, ..., M.

[0047] S32: Calculate the cost function value and weight of each particle in the M populations respectively. The specific calculation method is as follows:

[0048] (1) Calculate the cost function value of the i-th particle in the s-th population at time k i=1,2,......,N,s=1,2,......,M,k=1,2,......,T,the cost function is calculated as follows:

[0049]

[0050] Among them, 0<λ<1 is the forgetting factor, is the cost function value of the i-th particle in the s-th population at time k-1, is the i-th particle in the s-th population at time k The increment of is the increment function, y k is the particle observation value at time k, h(·) is the observation function, whose value is equal to the particle observation value at time k-1, and q≥1.

[0051] (2) Calculate the weight of the i-th particle in the s-th population at time k The weights are calculated as follows:

[0052]

[0053] Among them, β>0.

[0054] S33: Intelligent collaborative resampling of all particles in the M populations.

[0055] There are N particles in each population and the populations are independent of each other. The specific process of intelligent collaborative resampling of all particles in M ​​populations is as follows:

[0056] (1) Sort the particles in each population according to their weight from large to small, and divide the particles in each population into high-weight particles, medium-weight particles, and low-weight particles.

[0057] Figure 4 This is a diagram of the multi-population intelligent collaborative resampling mechanism, such as Figure 4 As shown in the figure, the particles in each population are sorted from large to small according to the weight, and then the particles in each population are divided into three parts according to the golden ratio, and the first 0.382N particles with larger weights are divided into high-weight particles. The last 0.382N particles with smaller weights are classified as low-weight particles The remaining particles with weights in the middle are classified as medium-weight particles. s=1,2,......,M。

[0058] (2) Retain high-weight particles and medium-weight particles during resampling.

[0059] Since high-weight particles With medium weight particles All of them are valid particles. In order to maintain the diversity of particles and improve the prediction accuracy of the model, high-weight particles with larger weights are directly retained during resampling. With medium weight particles

[0060] (3) An intelligent collaborative strategy is used to resample all particles in the low-weight particle set.

[0061] Low-weight particles with smaller weights The impact on the prediction results is small, so the strategy of intelligent collaboration is adopted All particles in the collection are resampled. Figure 4 As shown in the figure, in order to ensure the order of intelligent resampling, the resampling process between each population adopts a collaborative strategy based on a ring structure, that is, when resampling the low-weight particles in the sth population, the high-weight particles in the s-1th population are selected for collaboration, and when s=1, the high-weight particles in the Mth population are selected for collaborative resampling. However, when performing collaborative operations, the high-weight particles of the previous population cannot be directly copied to replace the low-weight particles of the current population like traditional resampling. This operation will cause a large number of particles with the same weight to be copied, and the interaction ability of particle information between different populations will decrease, resulting in a decrease in particle diversity and a decrease in the state assessment accuracy of the model.

[0062] In order to improve the diversity of particles among different populations and prevent blind copying of high-weight particles, when resampling all particles in the low-weight particle set, it is necessary to add an adaptive t-distribution mutation strategy based on the inter-population coordination of the above-mentioned ring structure. The main operation method is: randomly select a particle from the high-weight particle set of the s-1th population and perform adaptive t-distribution mutation, and replace the low-weight particle of the sth population with the mutated particle; when s=1, randomly select a particle from the high-weight particle set of M populations and perform adaptive t-distribution mutation, and replace the low-weight particle of the sth population with the mutated particle, s=1,2,......,M.

[0063] Adaptive t-distribution mutation is one of the common mutation methods of genetic algorithms. The effect of t-distribution mutation is between Gaussian mutation and Cauchy mutation. It combines the advantages of both and has good global search capabilities and local search capabilities in key areas.

[0064] A particle is randomly selected from the high-weight particle set of the s-1th population and adaptively mutated by t distribution, and the mutated particle is used to replace the low-weight particle of the sth population. The specific calculation method is as follows:

[0065]

[0066] in, is the lth low-weight particle of the sth population at time k after mutation and replacement, is the number of low-weight particles in the sth population at time k, is a high-weight particle randomly selected from the s-1th population at time k, is the number of high-weight particles in the s-1th population at time k, and q is the mutation rate.

[0067] When s=1, a particle is randomly selected from the high-weight particle set of M populations and adaptively mutated by t distribution, and the mutated particle is used to replace the low-weight particle of the sth population. The specific calculation method is as follows:

[0068]

[0069] in, is the lth low-weight particle of the first population at time k after mutation and replacement, is the number of low-weight particles in the first population at time k, is a high-weight particle randomly selected from the Mth population at time k, is the number of high-weight particles in the Mth population at time k, and q is the mutation rate.

[0070] After performing mutation operations on high-weight particles in each population to replace low-weight particles in the next population, and performing mutation operations on high-weight particles in the Mth population to replace low-weight particles in the first population, the high-weight particle set, medium-weight particle set, and low-weight particle set after intelligent resampling in each population are merged into a new population, while the particles between each population remain independent. The particle set between each population after resampling is as follows:

[0071]

[0072] Where s=1,2,......,M, M is the population size.

[0073] S34: Update the particles of each population in turn, and calculate the cost and weight of the updated particles. The specific calculation method is as follows:

[0074] (1) Each particle in the population is updated according to formula (5):

[0075]

[0076] in, is the i-th particle in the s-th population at time k after update, is the i-th particle in the s-th population at time k after resampling, The mean is The covariance matrix is ​​a Gaussian distribution with Σ, which can be set according to different usage scenarios.

[0077] (2) The particle cost of each population is updated according to formula (6):

[0078]

[0079] in, is the cost function value of the i-th particle in the s-th population at time k after the update, 0<λ<1 is the forgetting factor, is the cost function value of the i-th particle in the s-th population at time k-1 after the update, y k is the particle observation value at time k, h(·) is the observation function, whose value is equal to the particle observation value at time k-1, and q≥1.

[0080] The updated particle-cost set is

[0081] (3) After the update, the particle weight of each population is calculated according to formula (7):

[0082]

[0083] Among them, w ik s is the weight of the i-th particle in the s-th population at time k after update, β>0.

[0084] S35: Calculate the updated state estimation value of each population in turn, and calculate the state prediction value of the inverter within the specified time period according to the state estimation value of each population. The specific calculation method is as follows:

[0085] (1) Calculate the state estimate of each population at time k after the update according to formula (8):

[0086]

[0087] in, is the updated estimated state value of the sth population at time k, is the i-th particle in the s-th population at time k after update, is the weight of the i-th particle in the s-th population at time k after the update;

[0088] (2) Calculate the state prediction value of the inverter at time k according to formula (9):

[0089]

[0090] The k moment is specified, that is, the value of k is specified, and then the state prediction value of the inverter in the specified time period is calculated.

[0091] The data-driven inverter operating status assessment method provided by the embodiment of the present invention adopts a multi-population information interaction mechanism based on adaptive t-distribution variation in the process of predicting the operating status of the inverter using the inverter operating status prediction model based on multi-population collaborative CRPF, thereby enhancing the ability of particle information interaction between different populations, preventing a large number of high-weight particles from concentrating, improving sampling efficiency and particle diversity, and being able to effectively estimate the true posterior distribution, thereby improving the accuracy of inverter operating status assessment.

[0092] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0093] In one embodiment, a data-driven inverter operation status evaluation device is provided, and the data-driven inverter operation status evaluation device corresponds one-to-one to the data-driven inverter operation status evaluation method in the above embodiment. Figure 5As shown, the data-driven inverter operation status evaluation device includes a data acquisition module 10, a model building module 20, a state prediction module 30, an error calculation module 40, an error prediction module 50 and a state evaluation module 60. The functional modules are described in detail as follows:

[0094] The data acquisition module 10 is used to execute step S10 in the data-driven inverter operation status evaluation method in the above embodiment, that is, to acquire the status data of the inverter during operation.

[0095] The model building module 20 is used to execute step S20 in the data-driven inverter operation status assessment method in the above embodiment, that is, to build an inverter operation status prediction model based on multi-population collaborative CRPF based on status data, and to build an inverter operation status error correction model based on LSTM.

[0096] The state prediction module 30 is used to execute step S30 in the data-driven inverter operation state evaluation method in the above embodiment, that is, to calculate the state prediction value of the inverter within a specified time period using the inverter operation state prediction model based on multi-population collaborative CRPF.

[0097] The error calculation module 40 is used to execute step S40 in the data-driven inverter operation status evaluation method in the above embodiment, that is, to calculate the error between the state prediction value and the actual state value of the inverter within a specified time period.

[0098] The error prediction module 50 is used to execute step S50 in the data-driven inverter operation status evaluation method in the above embodiment, that is, to input the error into the LSTM-based inverter operation status error correction model to calculate the error prediction value.

[0099] The state evaluation module 60 is used to execute step S60 in the data-driven inverter operation state evaluation method in the above embodiment, that is, to calculate the final state prediction value according to the state prediction value and the error prediction value, and evaluate the operation state of the inverter according to the final state prediction value.

[0100] The data-driven inverter operation status evaluation device provided in the embodiment of the present invention obtains the status data of the inverter during operation, relies on the status data to construct an inverter operation status prediction model based on multi-population collaborative CRPF to predict the state prediction value of the inverter within a specified time period, and relies on the status data to construct an inverter operation status error correction model based on LSTM to predict the error between the state prediction value and the actual state value of the inverter within the specified time period to obtain an error prediction value, and finally calculates the final state prediction value according to the state prediction value and the error prediction value, and evaluates the operation status of the inverter according to the final state prediction value, thereby realizing real-time monitoring and evaluation of the inverter operation status, timely discovering the abnormal operation status of the inverter, improving the stability of the inverter during operation, reducing maintenance costs, and improving power quality and related economic benefits.

[0101] Compared with the existing inverter operation status assessment method, the data-driven inverter operation status assessment method provided by the embodiment of the present invention simultaneously utilizes two models, namely, the inverter operation status prediction model based on multi-population collaborative CRPF and the inverter operation status error correction model based on LSTM, to predict and deduce the operation status of the inverter, integrating the advantages of the two models. Among them, the inverter operation status prediction model based on multi-population collaborative CRPF can realize the prediction of the inverter operation status without a large amount of data, and has good prediction accuracy. The inverter operation status error correction model based on LSTM has a gating mechanism and a long-term memory mechanism, which can smoothly capture the interdependence of the state data during the operation of the inverter. The combined use of the two models not only overcomes the problem of low effectiveness of the state assessment caused by the small amount of historical data during the initial operation of the inverter, but also retains the correlation of the inverter operation status data, further improving the accuracy of the inverter operation status assessment. In addition, in the process of predicting the operating status of the inverter using the inverter operating status prediction model based on multi-population collaborative CRPF, a multi-population information interaction mechanism based on adaptive t-distribution variation is adopted to enhance the ability of particle information interaction between different populations and prevent the large concentration of high-weight particles. The sampling efficiency and particle diversity are improved, and the true posterior distribution can be effectively estimated, thereby improving the accuracy of inverter operating status evaluation.

[0102] For the specific limitations of the data-driven inverter operating status assessment device, please refer to the limitations of the data-driven inverter operating status assessment method above, which will not be repeated here. Each module in the above-mentioned data-driven inverter operating status assessment device can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0103] In one embodiment, a computer device is provided, whose internal structure diagram can be as follows: Figure 6 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with other external electronic devices via a network connection. When the computer program is executed by the processor, the functions or steps of the above-mentioned data-driven inverter operation status evaluation method are implemented.

[0104] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0105] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0106] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A data-driven inverter operating status evaluation method, characterized in that: include: Obtain status data of the inverter during operation; Based on the state data, an inverter operation state prediction model based on multi-population collaborative CRPF is constructed, and an inverter operation state error correction model based on LSTM is constructed; The inverter operation state prediction model based on the multi-population coordinated CRPF is used to calculate the state prediction value of the inverter within a specified time period; Calculate the error between the predicted state value and the actual state value of the inverter within a specified time period; Inputting the error into the inverter operation state error correction model based on LSTM to calculate the error prediction value; A final state prediction value is calculated according to the state prediction value and the error prediction value, and the operating state of the inverter is evaluated according to the final state prediction value.

2. The data-driven inverter operating status evaluation method according to claim 1, characterized in that: The inverter operation state prediction model based on multi-population coordinated CRPF is used to calculate the state prediction value of the inverter within a specified time period, including: Based on the state data, M populations are established, and the number of particles in each population is N; Calculate the cost function value and weight of each particle in the M populations respectively; Intelligent collaborative resampling of all particles in the M populations; Update the particles of each population in turn, and calculate the cost and weight of the updated particles; The updated state estimation value of each population is calculated in turn, and the state prediction value of the inverter within a specified time period is calculated based on the state estimation value of each population.

3. The data-driven inverter operating status evaluation method according to claim 2, characterized in that: The cost function value and weight of each particle in the M populations are calculated separately, including: Calculate the cost function value C of the i-th particle in the s-th population at time k k is , i=1,2,......,N,s=1,2,......,M,k=1,2,......,T,the cost function is calculated as follows: Among them, 0<λ<1 is the forgetting factor, is the cost function value of the i-th particle in the s-th population at time k-1, is the i-th particle in the s-th population at time k The increment of is the increment function, y k is the particle observation value at time k, h(·) is the observation function, whose value is equal to the particle observation value at time k-1, q≥1; Calculate the weight of the i-th particle in the s-th population at time k The weights are calculated as follows: Among them, β>0.

4. The data-driven inverter operating status evaluation method according to claim 2, characterized in that: The intelligent collaborative resampling of all particles in the M populations includes: Sort the particles in each population in descending order according to their weights, and divide the particles in each population into high-weight particles, medium-weight particles, and low-weight particles; Keep high-weight particles and medium-weight particles during resampling; An intelligent collaborative strategy is adopted to resample all particles in the low-weight particle set.

5. The data-driven inverter operating status evaluation method according to claim 4, characterized in that: The intelligent collaborative strategy is used to resample all particles in the low-weight particle set, including: Randomly extract particles from the high-weight particle set of the s-1th population and perform adaptive t-distribution mutation, and use the mutated particles to replace the low-weight particles of the sth population, s = 1, 2, ..., M; When s=1, particles are randomly selected from the high-weight particle set of M populations and adaptively mutated by t distribution, and the mutated particles are used to replace the low-weight particles of the sth population.

6. The data-driven inverter operating status evaluation method according to claim 2, characterized in that: The updating of particles of each population in turn and calculation of the cost and weight of the updated particles include: The particles in each population are updated according to formula (5): in, is the i-th particle in the s-th population at time k after update, is the i-th particle in the s-th population at time k after resampling, The mean is Gaussian distribution with covariance matrix Σ; The particle cost of each population is updated according to formula (6): in, is the cost function value of the i-th particle in the s-th population at time k after the update, 0<λ<1 is the forgetting factor, is the cost function value of the i-th particle in the s-th population at time k-1 after the update, y k is the particle observation value at time k, h(·) is the observation function, whose value is equal to the particle observation value at time k-1, q≥1; After updating, the particle weight of each population is calculated according to formula (7): in, is the weight of the i-th particle in the s-th population at time k after update, β>

0.

7. The data-driven inverter operating status evaluation method according to claim 2, characterized in that: The step of sequentially calculating the updated estimated state value of each population, and calculating the predicted state value of the inverter within a specified time period according to the estimated state value of each population, includes: According to formula (8), the state estimate of each population at time k after update is calculated: in, is the updated estimated state value of the sth population at time k, is the i-th particle in the s-th population at time k after update, is the weight of the i-th particle in the s-th population at time k after the update; According to formula (9), the state prediction value of the inverter at time k is calculated: The k moment is specified to obtain the state prediction value of the inverter within the specified time period.

8. The data-driven inverter operating status evaluation method according to any one of claims 1 to 7, characterized in that: The step of inputting the error into the inverter operation state error correction model based on LSTM to calculate the error prediction value includes: It is determined whether the error meets the error requirement. If not, the error is input into the LSTM-based inverter operation state error correction model to calculate the error prediction value.

9. A data-driven inverter operation status evaluation device, characterized in that: include: A data acquisition module is used to acquire the status data of the inverter during operation; A model building module, used to build an inverter operation state prediction model based on multi-population coordinated CRPF based on the state data, and to build an inverter operation state error correction model based on LSTM; A state prediction module, used to calculate the state prediction value of the inverter within a specified time period using the inverter operation state prediction model based on the multi-population collaborative CRPF; An error calculation module is used to calculate the error between the state prediction value and the actual state value of the inverter within a specified time period; An error prediction module, used for inputting the error into the inverter operation state error correction model based on LSTM to calculate an error prediction value; The state evaluation module is used to calculate a final state prediction value according to the state prediction value and the error prediction value, and evaluate the operating state of the inverter according to the final state prediction value.

10. A computer device, characterized in that: The computer device includes at least one processor and at least one memory communicatively connected to the processor, wherein the memory stores program instructions, and when the program instructions are called and executed by the processor, the data-driven inverter operating status assessment method according to any one of claims 1 to 8 is implemented.

Images