Photovoltaic power station health state detection method based on power generation capacity evaluation

Abstract

The invention relates to the technical field of photovoltaic power generation, and discloses a photovoltaic power station health state detection method based on power generation capacity evaluation, which is characterized in that four modules, namely a data acquisition module, a data cleaning module, a data set division and model training evaluation module and an evaluation result application module, are taken as cores. According to the photovoltaic power station health state detection method based on power generation evaluation, through cleaning and processing of historical data and combination of the Bayesian network model, the power generation health state of the power station and whether the power generation health state accords with expectation can be accurately evaluated, so that a reliable basis is provided for operation and maintenance of the power station, and the Bayesian model has unique advantages and is suitable for popularization and application. The method can make full use of the powerful probabilistic reasoning capability, deeply excavates the potential relation between the radiation quantity and the power generation capacity, achieves efficient and accurate power generation capacity evaluation, is especially suitable for the scene of power generation capacity evaluation depending on a single key environment factor, and has a wide application prospect. And a scientific quantitative evaluation method is provided for daily maintenance and fault diagnosis of the photovoltaic power station.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation, and in particular to a method for detecting the health status of a photovoltaic power station based on power generation evaluation. Background Art

[0002] As my country's demand for renewable energy and environmental protection increases, photovoltaic power generation and the new energy industry have become important forces in green development. As a clean energy source, photovoltaic power generation not only reduces fossil fuel consumption but also significantly reduces greenhouse gas emissions. However, its efficiency and economic benefits are affected by factors such as weather, equipment performance, and operation and maintenance management.

[0003] Although photovoltaic power generation technology has been widely adopted, it still faces challenges in power generation efficiency assessment, equipment monitoring, and fault warning. Traditional monitoring methods rely on manual inspections, which are inefficient and difficult to detect problems in a timely manner, affecting system operation and economic benefits.

[0004] Applying data cleaning technology and machine learning algorithms to photovoltaic power generation can accurately assess power generation efficiency and monitor equipment health in real time. Intelligent management can promptly handle abnormal situations, improve system reliability and economic efficiency, and promote the digital transformation of the industry. Summary of the Invention

[0005] (1) Technical problems solved In response to the shortcomings of the existing technology, the present invention provides a photovoltaic power station health status detection method based on power generation assessment, which provides support for photovoltaic system operation and maintenance, improves efficiency and benefits, and contributes to the construction of a green, low-carbon and sustainable energy system.

[0006] (2) Technical solution To achieve the above objectives, the present invention provides the following technical solution: a photovoltaic power station health status detection method based on power generation assessment, characterized by four core modules: data collection, data cleaning, data set partitioning and model training and evaluation, and application of evaluation results, and the following steps: Step 1: Data collection: Obtain key data such as power generation, radiation, and total installed capacity within a specified time period from the PV power station database; Step 2: Data cleaning: pre-process the acquired data, including deduplication, outlier detection, and sorting. At the same time, the data is cleaned using the calculation formula for the comprehensive efficiency of power generation, K, retaining only valid data with a K value between 0.5 and 1.0 and the corresponding date. Step 3: Dataset Partitioning and Model Training Evaluation: The cleaned dataset is clearly divided into a training set and a validation set. Based on a Bayesian random network, a one-dimensional input daily radiation and output power generation evaluation model is constructed. The training set is used for model training. Next, the validation set is used to evaluate the performance of the trained model to ensure the accuracy of the evaluation results. Step 4: Apply the evaluation results. Design and implement a visualization module for evaluating power generation. At the same time, debug the model prediction interface so that users can customize the input radiation and use the trained model to quickly obtain the corresponding evaluated power generation.

[0007] Step 5: System integration and packaging: effectively package the trained evaluation model and data cleaning method and integrate them into the system platform for practical application.

[0008] Preferably, the specific cleaning method corresponding to the step 2 "calculation formula for comprehensive power generation efficiency K" is:

[0009] In the formula: ----On-grid power generation (kW·h); ----Total solar radiation on the horizontal surface (kW·h / m2, peak hours); ---Installation capacity of components (kWp); ---Irradiance under standard conditions (constant = 1kW h / m2); ----Comprehensive efficiency coefficient; Step 1: Based on the formula for calculating power generation, derive the calculation formula for comprehensive efficiency K.

[0010] Step 2: Substitute the power generation, radiation and installed capacity data after slicing and summing operations into formula (1) to obtain the comprehensive efficiency K for each day in two years; Step 3: Set the judgment threshold based on the experience value. That is, the installed capacity varies, so the threshold set in the experiment here is set to 0.5-1.0 according to the specific installed capacity of the experiment. The data corresponding to the dates outside this range are deleted, and the data that meets the requirements are retained and stored in the new database as a cleaned high-quality data source.

[0011] Preferably, the implementation method of step 3 "data set division and model training evaluation" is: Step 1: Organize the cleaned power generation and radiation data and build the corresponding data set; The dataset consists of radiation and power generation. Each daily radiation and power generation corresponds to a piece of data. The data is read from the database where the cleaned data is stored. The high-quality data is divided into a training dataset and a validation dataset in a ratio of 7:3. Step 2: Build a Bayesian random network model; After the data is divided, the training set needs to be processed through a series of processes such as data normalization before being imported into the Bayesian random network model. The Bayesian random network can effectively capture the nonlinear relationship between radiation and power generation, while also handling the uncertainty in the input data. The basic idea of ​​Bayesian random networks is to express the conditional dependency between random variables through a directed acyclic graph (DAG). In this scheme, the input variable is the cumulative radiation, and the output variable is the cumulative power generation. The Bayesian random network implements variational inference by introducing random dropout during forward propagation, which means that during each forward propagation, a part of the neurons in the network will be dropped with a certain probability. is temporarily "discarded"; this strategy effectively simulates the randomness in the network structure, and then gives the probability distribution of power generation evaluation at the output layer; First, enter the radiation amount After the first hidden layer is processed, the activation function of the hidden layer is ReLU, and each neuron is The probability of random inactivation dropout:

[0012]

[0013] in, Is a Random variables of the same shape, each element is The probability of is 0; Represents element-wise multiplication, also called Hadamard product; Is a randomly generated binary vector, in which each element is 0 or 1, indicating whether the neuron at that position is randomly discarded (dropout); Operation, Those corresponding to The elements at the position with a median value of 0 will be "discarded" (i.e. set to 0), while the values ​​at other positions remain unchanged; this operation is the core step of dropout, which reduces overfitting and introduces randomness by randomly discarding some neurons; The random dropout operation introduces uncertainty into the output of each layer of neurons; next, the processed hidden layer Will be passed to the second hidden layer:

[0014]

[0015] Through this random inactivation operation, the network can effectively prevent overfitting during training, and also provide a better uncertainty estimate for the model; finally, the output layer gives the predicted power generation The probability distribution of :

[0016] in, is the average power generation predicted by the network, It represents the uncertainty of the prediction, that is, the variance of the power generation forecast value; that is, given , output Is subject to the mean , the variance is Normal distribution; optimize network weight parameters through maximum a posteriori estimation in Bayesian network To maximize the likelihood estimate of predicted power generation:

[0017] Random dropout is introduced to reduce overfitting of the model, and the Bayesian framework brings additional randomness and uncertainty quantification to each inference; The loss function consists of two parts: power generation estimation error and uncertainty estimation loss; by optimizing the following loss function, the network is able to balance estimation accuracy and uncertainty quantification:

[0018] By optimizing this loss function, the Bayesian random network can effectively learn the relationship between radiation and power generation, providing a more reliable power generation assessment.

[0019] The third step is to use the constructed dataset to train the model; The Bayesian random network built in the second step is trained on the dataset constructed in the first step. The number of iterations (epochs) is set to 3000 to ensure that the model can fully learn the data features. At the same time, the ReLU activation function is selected to enhance the nonlinear expression ability of the model. The training results will be evaluated based on the output result image and MSE (mean square error) indicator to determine the model's fitting effect and evaluation ability.

[0020] Preferably, step 4: result evaluation and error analysis; Use the model weight file trained in the third step and load the model to evaluate the data in random time periods in the validation set. In order to quantify the generation effect of the evaluation model, the mean square error (MSE) is used as the indicator of error analysis. MSE calculates the mean of the sum of squares of the errors of the corresponding points of the predicted data and the original data. The formula is:

[0021] in It's real data. is the data generated by the model, is an optional weight coefficient, n is the number of samples, and the smaller the MSE value, the closer the model result is to the actual data, the better the model fitting effect is, and the more successful the data evaluation is. This method can be used to objectively and accurately evaluate the evaluation ability of the model.

[0022] Preferably, during the data cleaning process, special attention is paid to the key indicator of "daily comprehensive efficiency K." The raw data is screened and filtered by determining whether the k value is within the normal range. Abnormal data outside the normal range is eliminated, and only normal dates and their corresponding data are retained to ensure the quality of the input data provided to the evaluation model. In the data cleaning rules, the judgment criteria for daily comprehensive efficiency K are set based on in-depth analysis of historical data and understanding of actual power generation conditions. Using the cleaned radiation data as the only data, a Bayesian network is applied to power generation assessment.

[0023] (3) Beneficial effects Compared with the existing technology, the present invention provides a photovoltaic power station health status detection method based on power generation evaluation, which has the following beneficial effects: This photovoltaic power station health status detection method based on power generation evaluation can accurately evaluate the health status of the power station's power generation and whether it meets expectations by cleaning and processing historical data and combining it with the Bayesian network model, thereby providing a reliable basis for the operation and maintenance of the power station. In addition, the Bayesian model has a unique advantage. It can grasp the probability distribution relationship between the radiation and power generation of the photovoltaic power station under normal operation by learning the distribution of data. When the power generation calculated by the model based on the radiation (assessment of power generation, ), it represents the power generation of the power station under normal operating conditions. We will evaluate the power generation and the actual power generation ( ) is defined as the health index ( ), once the health index falls below the preset threshold, it indicates that the power station may have an abnormality or fault; conversely, if the health index rises, it indicates that the operating status of the power station has improved, providing a scientific quantitative assessment method for daily maintenance and fault diagnosis of photovoltaic power stations. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a visualization diagram of the health assessment results of a photovoltaic power station health status detection method based on power generation assessment proposed by the present invention; Figure 2 This is a schematic diagram of the composition of a photovoltaic power generation intelligent assessment system based on a photovoltaic power station health status detection method based on power generation assessment proposed in the present invention; Figure 3 This is a schematic diagram of the Bayesian network structure of a photovoltaic power station health status detection method based on power generation assessment proposed by the present invention. DETAILED DESCRIPTION

[0025] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only 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 making creative efforts are within the scope of protection of the present invention.

[0026] Example 1: A photovoltaic power station health status detection method based on power generation assessment is characterized by comprising four core modules: data collection, data cleaning, data set division and model training and evaluation, and application of evaluation results, and includes the following steps: Step 1: Data collection: Obtain key data such as power generation, radiation, and total installed capacity within a specified time period from the PV power station database; Step 2: Data cleaning: Preprocess the acquired data, including deduplication, outlier detection, and sorting. At the same time, use the formula for calculating the comprehensive efficiency of power generation, K, to clean the data, retaining only valid data with K values ​​between 0.5 and 1.0 and the corresponding date. The specific cleaning method corresponding to the "comprehensive efficiency of power generation, K" formula is as follows:

[0027] In the formula: ----On-grid power generation (kW·h); ----Total solar radiation on the horizontal surface (kW·h / m2, peak hours); ---Installation capacity of components (kWp); ---Irradiance under standard conditions (constant = 1kW h / m2); ----Comprehensive efficiency coefficient; Step 1: Based on the formula for calculating power generation, derive the calculation formula for comprehensive efficiency K.

[0028] Step 2: Substitute the power generation, radiation and installed capacity data after slicing and summing operations into formula (1) to obtain the comprehensive efficiency K for each day in two years; Step 3: Set the judgment threshold based on the experience value. That is, the installed capacity varies, so the threshold set in the experiment here is set to 0.5-1.0 according to the specific installed capacity of the experiment. The data corresponding to the dates outside this range are deleted, and the data that meets the requirements are retained and stored in the new database as a cleaned high-quality data source.

[0029] Step 3: Dataset Partitioning and Model Training Evaluation: The cleaned dataset is clearly divided into a training set and a validation set. Based on a Bayesian random network, a one-dimensional input daily radiation and output power generation evaluation model is constructed. The training set is used for model training. Then, the validation set is used to comprehensively evaluate the performance of the trained model to ensure the accuracy of the evaluation results. Implementation method of step 3 "dataset partitioning and model training evaluation": Step 1: Organize the cleaned power generation and radiation data and build the corresponding data set; The dataset consists of radiation and power generation. Each daily radiation and power generation corresponds to a piece of data. The data is read from the database where the cleaned data is stored. The high-quality data is divided into a training dataset and a validation dataset in a ratio of 7:3. Step 2: Build a Bayesian random network model; After the data is divided, the training set needs to be processed through a series of processes such as data normalization before being imported into the Bayesian random network model. The Bayesian random network can effectively capture the nonlinear relationship between radiation and power generation, while also handling the uncertainty in the input data. The basic idea of ​​Bayesian random networks is to express the conditional dependency between random variables through a directed acyclic graph (DAG). In this scheme, the input variable is the cumulative radiation, and the output variable is the cumulative power generation. The Bayesian random network implements variational inference by introducing random dropout during forward propagation, which means that during each forward propagation, a part of the neurons in the network will be dropped with a certain probability. is temporarily "discarded"; this strategy effectively simulates the randomness in the network structure, and then gives the probability distribution of power generation evaluation at the output layer; First, enter the radiation amount After the first hidden layer is processed, the activation function of the hidden layer is ReLU, and each neuron is The probability of random inactivation dropout:

[0030]

[0031] in, Is a Random variables of the same shape, each element is The probability of is 0; Represents element-wise multiplication, also called Hadamard product; Is a randomly generated binary vector, in which each element is 0 or 1, indicating whether the neuron at that position is randomly discarded (dropout); Operation, Those corresponding to The elements at the position with a median value of 0 will be "discarded" (i.e. set to 0), while the values ​​at other positions remain unchanged; this operation is the core step of dropout, which reduces overfitting and introduces randomness by randomly discarding some neurons; The random dropout operation introduces uncertainty into the output of each layer of neurons; next, the processed hidden layer Will be passed to the second hidden layer:

[0032]

[0033] Through this random inactivation operation, the network can effectively prevent overfitting during training, and also provide a better uncertainty estimate for the model; finally, the output layer gives the predicted power generation The probability distribution of :

[0034] in, is the average power generation predicted by the network, It represents the uncertainty of the prediction, that is, the variance of the power generation forecast value; that is, given , output Is subject to the mean , the variance is Normal distribution; optimize network weight parameters through maximum a posteriori estimation in Bayesian network To maximize the likelihood estimate of predicted power generation:

[0035] Random dropout is introduced to reduce overfitting of the model, and the Bayesian framework brings additional randomness and uncertainty quantification to each inference; The loss function consists of two parts: power generation estimation error and uncertainty estimation loss; by optimizing the following loss function, the network is able to balance estimation accuracy and uncertainty quantification:

[0036] By optimizing this loss function, the Bayesian random network can effectively learn the relationship between radiation and power generation, providing a more reliable power generation assessment. Step 3: Use the constructed dataset to train the model; The Bayesian random network built in the second step is trained on the dataset constructed in the first step. The number of iterations (epochs) is set to 3000 to ensure that the model can fully learn the data features. At the same time, the ReLU activation function is selected to enhance the nonlinear expression ability of the model. The training results will be evaluated based on the output result image and MSE (mean square error) indicator to determine the model's fitting effect and evaluation ability.

[0037] Step 4: Apply the evaluation results. Design and implement a visualization module for the estimated power generation. At the same time, debug the model prediction interface so that users can customize the input radiation and quickly obtain the corresponding estimated power generation using the trained model. Step 5: System integration and packaging: effectively package the trained evaluation model and data cleaning method and integrate them into the system platform for practical application.

[0038] It should be noted that, during the data cleaning process, special attention is paid to the key indicator of "daily comprehensive efficiency K". By judging whether the k value is within the normal range, the original data is screened and filtered. Abnormal data beyond the normal range will be eliminated, and only normal dates and their corresponding data will be retained to ensure the quality of input data provided to the evaluation model. In the data cleaning rules, the judgment standard of daily comprehensive efficiency K is set based on the in-depth analysis of historical data and the understanding of actual power generation conditions. The Bayesian network is used to evaluate power generation using the only data of the cleaned radiation.

[0039] The beneficial effects of the present invention are as follows: the photovoltaic power station health status detection method based on power generation evaluation can accurately evaluate the power generation health status of the power station and whether it meets expectations by cleaning and processing historical data and combining with the Bayesian network model, thereby providing a reliable basis for the operation and maintenance of the power station. Moreover, the Bayesian model has a unique advantage in that it can grasp the probability distribution relationship between the radiation and power generation of the photovoltaic power station under normal operation by learning the distribution of data. When the power generation calculated by the model based on the radiation (evaluation of power generation, ), it represents the power generation of the power station under normal operating conditions. We will evaluate the power generation and the actual power generation ( ) is defined as the health index ( ), once the health index falls below the preset threshold, it indicates that the power station may have an abnormality or fault; conversely, if the health index rises, it indicates that the operating status of the power station has improved, providing a scientific quantitative assessment method for daily maintenance and fault diagnosis of photovoltaic power stations.

[0040] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A photovoltaic power station health status detection method based on power generation evaluation, characterized in that: The four core modules are data collection, data cleaning, data set division and model training evaluation, and evaluation result application, including the following steps: Step 1: Data collection: Obtain key data such as power generation, radiation, and total installed capacity within a specified time period from the PV power station database; Step 2: Data cleaning: pre-process the acquired data, including deduplication, outlier detection, and sorting. At the same time, the data is cleaned using the calculation formula for the comprehensive efficiency of power generation, K, retaining only valid data with a K value between 0.5 and 1.0 and the corresponding date. Step 3: Dataset Partitioning and Model Training Evaluation: The cleaned dataset is clearly divided into a training set and a validation set. Based on a Bayesian random network, a one-dimensional evaluation model of input daily radiation and output power generation is constructed. The training set is used for model training. Next, the validation set is used to comprehensively evaluate the performance of the trained model to ensure the accuracy of the evaluation results. Step 4: Apply the evaluation results. Design and implement a visualization module for the estimated power generation. At the same time, debug the model prediction interface so that users can customize the input radiation and quickly obtain the corresponding estimated power generation using the trained model. Step 5: System integration and packaging: effectively package the trained evaluation model and data cleaning method and integrate them into the system platform for practical application.

2. A photovoltaic power station health status detection method based on power generation assessment according to claim 1, characterized in that: The specific cleaning method corresponding to the "calculation formula for comprehensive power generation efficiency K" in step 2 is as follows: In the formula: E P ----On-grid power generation (kW·h); H A ----Total solar radiation on the horizontal surface (kW·h / m2, peak hours); P AZ --- Module installed capacity (kWp); E S --- Irradiance under standard conditions (constant = 1kW h / m2); K----comprehensive efficiency coefficient; Step 1: Based on the formula for calculating power generation, derive the formula for calculating comprehensive efficiency K; Step 2: Substitute the power generation, radiation and installed capacity data after slicing and summing operations into formula (1) to obtain the comprehensive efficiency K for each day within the time range of the data set; Step 3: Set the judgment threshold according to the comprehensive efficiency index of the photovoltaic power station. AZ That is, the installed capacity varies, so the threshold set in the experiment here is set to 0.5-1.0 according to the specific installed capacity of the experiment. The data corresponding to the dates outside this range are deleted, and the data that meets the requirements are retained and stored in the new database as a cleaned high-quality data source.

3. The photovoltaic power station health status detection method based on power generation evaluation according to claim 1 is characterized in that: Implementation method of step 3 "dataset division and model training evaluation": Step 1: Organize the cleaned power generation and radiation data and build the corresponding data set; The dataset consists of radiation and power generation. Each daily radiation and power generation corresponds to a piece of data. The data is read from the database where the cleaned data is stored. The high-quality data is divided into a training dataset and a validation dataset in a ratio of 7:

3. Step 2: Build a Bayesian random network model; After the data is divided, the training set needs to be processed through a series of processes such as data normalization before being imported into the Bayesian random network model. The Bayesian random network can effectively capture the nonlinear relationship between radiation and power generation, while also handling the uncertainty in the input data. The Bayesian stochastic network implements variational inference by introducing random dropout during forward propagation. This means that during each forward propagation, a portion of the neurons in the network are temporarily "dropped" with a certain probability p. This strategy effectively simulates the randomness in the network structure and provides a probability distribution for power generation estimation at the output layer. First, the input radiation x is processed through the first hidden layer, the activation function of the hidden layer is ReLU, and each neuron is randomly inactivated with probability p: h1=ReLU(W1x+b1)#(2) drop(h1)=h1☉r1#(3) Here, r1 is a random variable with the same shape as h1, where each element is 0 with probability p. ⊙ represents element-wise multiplication, also called the Hadamard product. r1 is a randomly generated binary vector, where each element is either 0 or 1, indicating whether the neuron at that position is randomly dropped (dropout). Through the ⊙ operation, the elements in h1 corresponding to positions with a value of 0 in h1 are "dropped" (i.e., set to 0), while the values ​​at other positions remain unchanged. This operation is the core step of dropout, which reduces overfitting and introduces randomness by randomly dropping some neurons. At the output of each layer of neurons, the random dropout operation introduces uncertainty; next, this processed hidden layer h1 is passed to the second hidden layer: h2=ReLU(W2h1+b2)#(4) drop(h2)=h2☉r2#(5) Through this random inactivation operation, the network can effectively prevent overfitting during training, while also providing a better uncertainty estimate for the model. Ultimately, the output layer gives the probability distribution of the predicted power generation y: Among them, μ(x) is the mean power generation predicted by the network, σ 2 (x) represents the uncertainty of the prediction, that is, the variance of the power generation prediction value; that is, given x, the output y is subject to the mean μ(x) and the variance σ 2 (x) is normally distributed; by maximum a posteriori estimation in the Bayesian network, the network weight parameter W is optimized to maximize the likelihood estimation of the predicted power generation: Random dropout is introduced to reduce overfitting of the model, and the Bayesian framework brings additional randomness and uncertainty quantification to each inference; The loss function consists of two parts: power generation estimation error and uncertainty estimation loss; by optimizing the following loss function, the network is able to balance estimation accuracy and uncertainty quantification: By optimizing this loss function, the Bayesian random network can effectively learn the relationship between radiation and power generation, providing a more reliable power generation assessment. Step 3: Use the constructed dataset to train the model; The Bayesian random network built in the second step is trained on the dataset constructed in the first step. The number of iterations (epoch) is set to 3000 to ensure that the model can fully learn the data features. At the same time, the ReLU activation function is selected to enhance the nonlinear expression ability of the model. The training results will be evaluated based on the output result distribution image and MSE (mean square error) indicator to determine the model's fitting effect and evaluation ability.

4. The photovoltaic power station health status detection method based on power generation evaluation according to claim 1, characterized in that: Step 4: Result evaluation and error analysis; Use the model weight file trained in the third step and load the model. Set the target time period in the validation set to obtain data for evaluation. In order to quantify the generation effect of the evaluation model, the mean square error (MSE) will be used as the indicator of error analysis. MSE calculates the mean of the sum of squares of the errors of the corresponding points of the predicted data and the original data. The formula is: where y i It's real data. is the data generated by the model, w i is an optional weight coefficient, n is the number of samples, and the smaller the MSE value, the closer the model result is to the actual data, the better the model fitting effect is, and the more successful the data evaluation is. This method can be used to objectively and accurately evaluate the evaluation ability of the model.

5. The photovoltaic power station health status detection method based on power generation evaluation according to claim 1 is characterized in that: During the data cleaning process, special attention is paid to the key indicator of "daily comprehensive efficiency K". By judging whether the k value is within the normal range, the original data is screened and filtered. Abnormal data outside the normal range will be eliminated, and only normal dates and their corresponding data will be retained to ensure the quality of the input data provided to the evaluation model. In the data cleaning rules, the judgment standard of daily comprehensive efficiency K is set based on an in-depth analysis of historical data and an understanding of actual power generation conditions. The Bayesian network is used to evaluate power generation using the cleaned single-dimensional radiation data.

Images