A photovoltaic string power loss analysis and identification method and storage medium thereof

Abstract

This invention provides a photovoltaic string power loss analysis method. The method involves constructing a multidimensional dataset using multiple interrelated environmental variables and partitioning the dataset using a slicing or dicing method. The slicing or dicing method combines the optimal power generation under different environmental conditions with other variables to establish a power generation model. Real-time environmental variables and actual power generation data are collected, and the actual power generation is compared with the corresponding data of the environmental variables in the power generation model to determine the power loss. Based on the analysis method, an identification method is proposed to determine the power loss after manually eliminating the influence of non-periodic variables based on real-time data feedback. This invention solves the problems of inaccurate power loss data analysis and inaccurate problem identification.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation monitoring, and in particular to a method for analyzing and identifying photovoltaic string power loss. Background Art

[0002] With the increasing importance of clean energy in the energy mix and national energy security, photovoltaic power generation has always been a key industry supported by the state. Currently, my country's photovoltaic power generation technology is at an internationally leading level. In recent years, with the annual expansion of photovoltaic installed capacity, the amount of electricity connected to the grid has also gradually increased, and my country's photovoltaic industry has entered a stage of large-scale development. The power generation level of a photovoltaic power station during operation is the most critical factor directly affecting the economic benefits of a photovoltaic power station. This level of power generation is closely related to the daily maintenance of the power station, requiring timely inspection and maintenance of faulty equipment and equipment status. Photovoltaic power stations occupy large areas and have a large number of devices, making it difficult to detect faulty equipment through manual monitoring.

[0003] The power generation process of photovoltaic power stations is affected by weather changes, equipment wear and tear, and obstruction by foreign objects. Weather factors are uncertain and difficult to accurately predict, especially their long-term impact. Equipment wear and obstruction by foreign objects can be monitored and identified through data analysis and algorithmic models.

[0004] In terms of anomaly detection, most current market approaches rely on computer vision to identify PV panel anomalies. For example, foreign objects such as buildings, plants, and dust can cause inefficient power generation and power loss. However, using computer vision to identify obstructions makes it difficult to accurately quantify the extent of the loss. Furthermore, it is also difficult to determine the extent of dust accumulation on PV panels.

[0005] In terms of data analysis, the market currently primarily uses data from the power generation process to identify inefficient components and quantify power losses. Current deviation calculations are typically based on string current data, using clustering methods such as K-nearest neighbor and K-means to identify abnormally fluctuating string currents. Thresholds are then set based on the clustering results, the location of the PV plant, and empirical evidence. However, existing technologies only use clustering algorithms to cluster thresholds derived from different environmental conditions. When there are correlations between environmental conditions, threshold deviations can fluctuate.

[0006] The above two problems lead to inaccurate analysis of PV string power loss and inaccurate problem identification. Summary of the Invention

[0007] In view of the deficiencies in the prior art, the present invention provides a method for analyzing and identifying power loss in photovoltaic strings, which solves the problem in the prior art of inaccurate analysis and identification of power loss in photovoltaic strings.

[0008] To achieve the above purpose, the present invention adopts the following technical solutions: a photovoltaic string power generation loss analysis method,

[0009] The method includes the following steps: forming a multidimensional data set with multiple interrelated environmental variables, and dividing the data set by slicing or dicing; establishing a power generation model by combining the slices or dicing and other variables under different environmental conditions with the optimal power generation under the environmental conditions; collecting real-time environmental variables and actual power generation data, comparing the actual power generation with the corresponding data of the environmental variables in the model, and determining the power loss.

[0010] A method for identifying power generation loss in photovoltaic strings is also proposed, which uses the above-mentioned analysis method and also includes establishing a cleanliness index model based on the impact of different dust coverage levels on power generation loss; the optimal power generation is obtained when there is no dust accumulation on the surface of the component.

[0011] A storage medium for storing programs for running the above two methods is also proposed.

[0012] Compared with the prior art, the present invention has the following beneficial effects:

[0013] After the present invention forms a multidimensional data set with interrelated environmental variables, it uses slicing or dicing to correlate the multidimensional data and then comprehensively considers them, avoiding using only clustering algorithms to cluster the thresholds obtained under different environmental conditions, which causes the thresholds to fluctuate due to the mutual influence of interrelated conditions, thereby deriving a more reasonable power generation model; and then using the influence of dust coverage on power generation loss to establish a cleaning index model, it is possible to identify the dust coverage of the current equipment, which is convenient for subsequent maintenance processing.

[0014] Other advantages, objectives and features of the present invention will be reflected in part from the following description and will be understood by those skilled in the art through study and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A logic diagram is constructed for the power generation model of an embodiment of the present invention.

[0016] Figure 2 Schematic diagram of the recognition process of an embodiment of the present invention.

[0017] Figure 3 This is a statistical chart of power generation and irradiance data.

[0018] Figure 4 Statistical chart of power generation and ambient temperature data.

[0019] Figure 5 Statistical chart of power generation and wind speed data. DETAILED DESCRIPTION

[0020] In order to make the technical means, creative features, objectives and functions achieved by the present invention clearer and easier to understand, the technical solutions in the present invention are further described below with reference to the accompanying drawings and embodiments.

[0021] like Figure 1 and Figure 2 As shown, an embodiment of the present invention proposes a method for analyzing photovoltaic string power generation loss, including the following steps: using multiple interrelated different environmental variables to form a multidimensional data set, and dividing the data set by slicing or dicing; the slicing or dicing algorithm can be flexibly selected according to the nature of the data set.

[0022] The slices or blocks and other variables under different environmental conditions are combined with the optimal power generation under those environmental conditions to establish a power generation model. Statistical methods can be used to learn the optimal power generation performance of the components and establish a power generation model.

[0023] Collect real-time environmental variables and actual power generation data, compare the actual power generation with the corresponding data of environmental variables in the model, and determine the power loss.

[0024] The main factors affecting power generation include time, irradiance, ambient temperature and wind speed; Figure 3 As shown in the figure, irradiance, as a physical quantity of solar radiation intensity, is the most direct meteorological factor that determines the output of photovoltaic modules. The greater the irradiance, the greater the photovoltaic power generation. Therefore, it can be considered that the output power of photovoltaic modules is basically proportional to the irradiance and that the irradiance and photovoltaic power generation are highly correlated. Figure 4 As shown in the figure, ambient temperature is a physical quantity that indicates the degree of air temperature. Its value reflects the change of irradiance to a certain extent. Within a certain range, the higher the temperature, the greater the photovoltaic power generation. Figure 5 As shown in the figure, theoretically, the increase in wind speed promotes the air flow on the surface of the solar cell module to a certain extent, thereby reducing the surface temperature of the module, which is beneficial to improving the photoelectric conversion efficiency and increasing the output power. However, as the wind speed increases, it has a negative impact on the irradiance, and the overall power generation power gradually decreases after a brief increase.

[0025] Therefore, irradiance, ambient temperature, and wind speed are interrelated environmental factors, and the impact of any one factor on power generation is influenced by the other two. Time, irradiance, ambient temperature, and wind speed are constructed into a multidimensional dataset, which is then partitioned using a slicing and dicing method. For each slice or dicing, dirty data is processed using data cleaning methods (such as the 3 sigma criterion, interpolation, and relational constraints). Statistical methods (such as the Gaussian model and maximum likelihood estimation) are then applied to learn the optimal power generation for each slice or dicing. This yields a power generation model that correlates the optimal power generation with time, irradiance, ambient temperature, and wind speed. Aggregating and slicing the associated data before incorporating it into the model avoids the problem of insufficient consideration of inter-data relationships when considering individual data.

[0026] The slicing and dicing method can adopt the algorithms in the existing technology. It can also construct a coordinate system with the irradiance, ambient temperature and wind speed coordinate axes, set a maximum and minimum value for each factor, and obtain a square block model. Then, the square is cut into small blocks at very small numerical intervals. Each block corresponds to an optimal power generation power, and the optimal power generation can be obtained by integrating it with time.

[0027] Based on this, the present invention also proposes a method for identifying power loss in photovoltaic strings. The optimal power generation is determined when there is no dust accumulation on the module surface. After constructing a power generation model, a cleanliness index model is developed based on a deep convolutional neural network to evaluate the degree of dust accumulation and quantify its impact on the module's power generation capacity. The degree of dust accumulation's impact on the module's power generation capacity (cleanliness index) is reflected in the difference between the optimal power generation and the actual power generation. Therefore, based on the power generation model and combined with the strong learning ability of the deep convolutional neural network, accurate identification of dust accumulation (cleanliness index) can be achieved. In addition to time, irradiance, ambient temperature, and wind speed as the primary calculation data for optimal power generation prediction, many other influencing factors exist, such as building shading, plant shading, cloud cover, dust accumulation, faults, and downtime. Some of these influences exhibit a certain regularity and are predictable, and are classified as periodic variables, including building shading and plant shading. Some influences, such as cloud cover, faults, and downtime, are time-invariant or difficult to predict, and are classified as non-periodic variables, but can be recorded through meteorological data and device feedback. A variation model can be set up to account for periodic variables. For example, in the short term, it's possible to predict which strings will be blocked by a fixed building at what time each day and to what extent. The resulting change in power generation is modeled relatively fixed. When power generation exhibits this kind of variation over time, it can be assumed that the blockage was caused by the building or vegetation, and the impact of dust is not accounted for. The impact of non-periodic variables is investigated and calculated using real-time weather data and equipment feedback. Ultimately, the impact of dust accumulation losses is determined.

[0028] After the abnormal influence is divided into periodic variables and non-periodic variables, some information that is difficult to be fed back by the equipment can be automatically excluded through the change model, while some information that can be fed back by the equipment or external data can be manually judged, making it possible to monitor and calculate the dust accumulation loss. Figure 2 The environment variables in do not include non-periodic environment variables.

[0029] The power generation model inputs time, actual power generation, radiation intensity, ambient temperature, and wind speed, combined with the inverter's operating status (fault, maintenance, power curtailment, or shutdown), calculates power generation lost due to dust accumulation. By analyzing and identifying power generation losses, we can gain a comprehensive understanding of controllable PV strings. Subsequently, cleaning and maintenance can be scheduled based on the cleanliness index and power generation lost due to dust accumulation.

[0030] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. 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 purpose 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 photovoltaic string power generation loss analysis method, characterized in that: The steps include: Use multiple interrelated and different environmental variables to form a multidimensional data set, and use slicing or dicing methods to divide the data set; The slices or blocks and other variables under different environmental conditions are combined with the optimal power generation under the environmental conditions to establish a power generation model; Collect real-time environmental variables and actual power generation data, compare the actual power generation with the corresponding data of environmental variables in the power generation model, and determine the power loss.

2. A photovoltaic string power generation loss analysis method according to claim 1, characterized in that: The environmental variables include three or any two of irradiance, ambient temperature and wind speed.

3. A photovoltaic string power generation loss analysis method according to claim 2, characterized in that: The other conditions include periodic variables and non-periodic variables, and the periodic variables, the slicing or dicing, and the optimal power generation are used together to establish a power generation model.

4. A method for identifying photovoltaic string power generation loss, characterized in that: The analysis method according to any one of claims 1 to 3 is used to determine the power loss after manually eliminating the influence of non-periodic variables based on real-time data feedback.

5. A method for identifying photovoltaic string power generation loss according to claim 4, characterized in that: A cleanliness index model is established by analyzing the effect of different dust coverage levels on power generation loss; the optimal power generation is obtained when there is no dust accumulation on the component surface.

6. A photovoltaic string power generation loss identification method according to claim 5, characterized in that: The periodic variables include building shading and plant shading.

7. A method for identifying photovoltaic string power generation loss according to claim 6, characterized in that: The non-periodic variables include cloud cover, failures and outages.

8. A computer storage medium, characterized in that The computer storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of the analysis method according to any one of claims 1 to 3.

9. A computer storage medium, characterized in that The computer storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of the identification method according to any one of claims 4 to 8.

Images