System and method for generation prediction, fault detection and performance enhancement of photovoltaic generating station

Abstract

The present invention relates to a system and method for predicting photovoltaic (PV) power generation, detecting faults, and enhancing the performance of PV generating stations The system comprises a data collection module (14) that acquires actual data on environmental conditions and PV system performance and transmits sensor data to a cloud platform for analysis, the data analysis module (15) processes data to predict PV power generation, optimize system performance, and identify potential issues, and user interface (16) display system performance, provide accurate understandings, and enable remote monitoring. The system and method utilize advanced machine learning techniques to improve the accuracy of PV power generation predictions, detect faults, and optimize system performance, resulting in increased energy production and reduced operational costs.

Description

[0001] “SYSTEM AND METHOD FOR GENERATION PREDICTION, FAULT DETECTION, AND PERFORMANCE ENHANCEMENT OF PHOTOVOLTAIC GENERATING STATION”

[0002] FIELD OF INVENTION:

[0003] The present invention relates to the field of operation and maintenance of photovoltaic (PV) generating stations. More particularly, the present invention relates to a system and method for generation prediction, fault detection, and performance enhancement of photovoltaic generating station which incorporates performance benchmarking, generation prediction, anomaly detection, and performance improvement modules thereby, utilizes advanced algorithms and realtime data analysis. The system aims to maximize energy output, minimize downtime, and enhance the overall efficiency of PV stations.

[0004] BACKGROUND OF THE INVENTION:

[0005] Photovoltaic (PV) generating stations have a huge number of generating units spread across a large area. With larger generating stations being installed, the difficulty in monitoring the performance of each generating unit increases manifold. A typical PV generating station consists of lakhs of modules, and to ensure that the station is optimally running, each module and its connected system need to be in a healthy state. Conventional PV generator performance monitoring methods are top-down, i.e., when any performance is detected at the aggregate level, then the analysis of string-level data is done to locate the issue. This whole process is laborious, time-consuming, and less reliable.

[0006] To resolve this problem, an autonomous system has been invented that predicts the expected performance at the string level and detects if the actual performance is within the expected range or not. This enables automated performance monitoring and effortless and quick detection of issues at high granularity in PV generating stations.

[0007] Further, it enables early detection of performance issues at a very granular level, which may go unnoticed for a long time with conventional operation and maintenance approaches. This results in significant savings with increased generation and longevity of the generating equipment.

[0008] So, in regard to the aforesaid conventional disadvantageous photovoltaic (PV) generating stations, further the idea of using an autonomous system that predicts string-level performance and detects if actual performance is within the expected range, enabling automated monitoring and quick detection of issues in PV generating stations. This early detection leads to significant savings and equipment longevity.

[0009] PRIOR ARTS AND THEIR DISADVANTAGES:

[0010] There are various inventions related to the present invention but none satisfies the objects of inventions laid down by the present invention:

[0011] The patent application number CN201510047536A discloses a photovoltaic power station with a control method. It consists of a photovoltaic (PV) generating device and a synchronous coordinator wirelessly communicating with each device. The inverter uses a first wireless control module and a DC-AC (Direct Current-Alternating Current) conversion module to convert DC current output into AC current synchronized with characteristic parameters. This reduces power loss and improves conversion efficiency of electric energy.

[0012] However, said prior art depicts a synchronous controller that measures the grid parameters and sends a control signal to inverter functioning. Hence, the cited art fails to disclose information about a system for autonomously optimizing photovoltaic (PV) generating stations that incorporates performance benchmarking, generation prediction, anomaly detection, performance improvement modules, utilizing advanced algorithms, and real-time data analysis. The cited prior art also fails to disclose about the system that maximizes energy output, minimizes downtime, and enhances the overall efficiency of PV systems.

[0013] The patent application number CN202111423040A discloses about a distributed photovoltaic operation data acquisition system and data processing method. It involves calculating theoretical and actual photovoltaic power generation, eliminating abnormal data, drawing a typical curve, performing fault pre-judgment, and determining an operation and maintenance scheme. The method improves fault judgment accuracy, provides a corresponding operation and maintenance scheme, and enables quick response and maintenance. The system also includes a fault occurrence moment calculation and analysis.

[0014] However, said prior art fails to efficiently use sensors and data pipelines to automate generation benchmarking and comparison and is not suited for the operation of utility-scale photovoltaic (PV) generating stations. Furthermore, the cited invention fails to suggest the uses of advanced data engineering and Al / ML (Artificial Intelligence / Machine Learning) techniques to automatically compare actual and predicted generation, flag deviations for maintenance action, and detect low- performance causes. Also, it fails to provide a highly robust and flexible PV generating monitoring system that can function well even without the requirement of PV panel specifications. Hence, the cited art fails to disclose information about a system for autonomously optimizing photovoltaic (PV) generating stations that incorporates performance benchmarking, generation prediction, anomaly detection, performance improvement modules, utilizing advanced algorithms, and real-time data analysis. The cited prior art also fails to disclose about the system that maximizes energy output, minimizes downtime, and enhances the overall efficiency of PV systems.

[0015] The patent application number CN201810365216A discloses a defect diagnosis method for photovoltaic power stations. It involves collecting test data, determining annual performance ratio (PR) values, and diagnosing layer-by-layer defects using a performance ratio (PR) evaluation threshold model. The threshold model is predetermined based on actual measured data from the power station. This method allows defect detection at the operational level, improving the efficiency of defect diagnosis and enhancing the overall efficiency of the power station.

[0016] However, said prior art fails to provide a system that relies on advanced Al / ML (Artificial Intelligence / Machine Learning) techniques to automatically compare actual and predicted generation, flag deviations for maintenance action, detect low-performance causes, and check if the actual performance is up to the mark or not. Also, it fails to use string-level data to check the performance. Hence, the cited art fails to disclose information about a system for autonomously optimizing photovoltaic (PV) generating stations that incorporates performance benchmarking, generation prediction, anomaly detection, performance improvement modules, utilizing advanced algorithms, and real-time data analysis. The cited prior art also fails to disclose about the system that maximizes energy output, minimizes downtime, and enhances the overall efficiency of PV systems.

[0017] DISADVANTAGES OF THE PRIOR ART: All prior arts suffer from all or at least any of the below-mentioned disadvantages :

[0018] • Many of the prior art fail to provide a fully autonomous solution for optimizing PV power generation.

[0019] • Many of the prior art rely on manual intervention or predefined rules, limiting their ability to adapt to changing conditions and maximize energy output.

[0020] • Many of the prior art fails to effectively utilize advanced data engineering and Al / ML techniques to extract valuable insights from real-time data.

[0021] • Many of the prior art lack the capability to perform comprehensive performance benchmarking, generation prediction, and anomaly detection, hindering their ability to identify and address performance issues.

[0022] • Many of the prior art systems often operate at a coarse-grained level and fail to monitor performance at the string or module level.

[0023] • Many of the prior art limits their ability to pinpoint specific areas of concern and take targeted corrective actions.

[0024] • Many of the prior art systems require specific hardware or configuration information, limiting their flexibility and adaptability to different PV installations, which may hinder their deployment and effectiveness in diverse environments.

[0025] • Many of the prior art describe systems that may not accurately detect and diagnose faults promptly, leading to increased downtime and reduced energy output.

[0026] • Many of the prior art may rely on manual inspection or predefined rules, which may be time-consuming and error-prone. OBJECTS OF THE PRESENT INVENTION:

[0027] The main object of the present invention is to provide a system and method for generation prediction, fault detection, and performance enhancement of photovoltaic generating station.

[0028] The principal object of the present invention is to provide a system and method for generation prediction, fault detection, and performance enhancement of photovoltaic generating station that eliminates redundant data points from historical station performance.

[0029] Another object of the present invention is to provide a system and method for generation prediction, fault detection, and performance enhancement of photovoltaic generating station that ensures that the golden dataset is continuously updated to incorporate changes in generating system configuration and parameters.

[0030] Yet another object of the present invention is to provide a system and method for generation prediction, fault detection, and performance enhancement of photovoltaic generating station that can be easily implemented across plants with different configurations or technologies.

[0031] Further object of the present invention is to provide a system and method for generation prediction, fault detection, and performance enhancement of photovoltaic generating station that establishes benchmark performance and identifies performance issues.

[0032] Another object of the present invention is to provide a system and method for generation prediction, fault detection, and performance enhancement of photovoltaic generating station that notifies site managers of detected issues.

[0033] Yet another object of the present invention is to provide a system and method for generation prediction, fault detection, and performance enhancement of photovoltaic generating station that has a provision to update corrective action taken and close raised tickets.

[0034] Another object of the present invention is to provide a system and method for generation prediction, fault detection, and performance enhancement of photovoltaic generating station with a user-friendly, intuitive graphical user interface on which various performance KPIs and savings can be visualized, enabling efficient plant monitoring and issue rectification.

[0035] Further object of the present invention is to provide a system and method for generation prediction, fault detection, and performance enhancement of photovoltaic generating station that enables early identification of the smallest performance issues in a PV generating station, resulting in significant cost savings and increasing plant longevity.

[0036] Yet another object of the present invention is to provide a system and method for generation prediction, fault detection, and performance enhancement of photovoltaic generating station that overcomes major problems of conventional methods.

[0037] BRIEF DESCRIPTION OF DRAWING:

[0038] Various other objects, features, and attendant advantages of the present invention will become fully appreciated as the same becomes better understood when considered in conjunction with the accompanying drawings, in which like reference characters designate the same or similar parts throughout the several views, and wherein:

[0039] SUMMARY OF THE INVENTION:

[0040] The present invention relates to a system and method for predicting photovoltaic (PV) power generation, detecting faults, and enhancing the performance of PV generating stations. The system comprises a data collection module that acquires actual data on environmental conditions and PV system performance and transmits sensor data to a cloud platform for analysis, the data analysis module processes data to predict PV power generation, optimize system performance, and identify potential issues, and user interface displays system performance, provides accurate understandings, and enable remote monitoring. The system and method utilize advanced machine learning techniques to improve the accuracy of PV power generation predictions, detect faults, and optimize system performance, resulting in increased energy production and reduced operational costs.

[0041] LIST OF REFERENCE NUMERALS

[0042] 1 : Photovoltaic module

[0043] 2: Pyranometer

[0044] 3: Temperature sensor

[0045] 4: Humidity sensor

[0046] 5: Anemometer

[0047] 6: Current sensor 7 : Voltage sensor

[0048] 9: Inverter

[0049] 10: String

[0050] 1 1 : Grid

[0051] 14: Data collection module

[0052] 15: Data analysis module

[0053] 16: User interface

[0054] DETAILED DESCRIPTION OF THE INVENTION:

[0055] The following description is presented to enable any person skilled in the art to make and use the invention. It is provided in the context of a particular application and its requirements. Various modifications to the disclosed embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments and applications without departing from the spirit and scope of the present invention. Thus, the present invention is not intended to be limited to the embodiments shown but is to be accorded the widest scope consistent with the principles and features disclosed herein.

[0056] It is to be understood that the term “comprising” or “comprises” used in the specification and claims refer to the element of the invention which comprises X, Y, and Z, which means that the invention might have other elements in addition to X, Y, and Z. For example, their invention could include A, B, and / or C as long as it also has X, Y, and Z. The present invention provides a system and method for generation prediction, fault detection, and performance enhancement of a photovoltaic generating station is a power plant that is installed in the site in which the system uses weather monitoring sensors, energy generation units, and data storage devices to predict photovoltaic (PV) string expected generation. Furthermore, said system uses advanced data engineering and Al / ML (Artificial Intelligence / Machine Learning) techniques to automatically compare actual and predicted generation, flag deviations for maintenance action, and detect low-performance causes.

[0057] Now referring to Figure 1, a photovoltaic (PV) generating station comprising:

[0058] • a plurality of PV panels (1) configured to convert incident radiation into direct current (DC) electrical energy. Each PV panel (1) includes a semiconductor material capable of absorbing photons from sunlight and generating an electric current,

[0059] • a plurality of inverters (9) configured to convert DC electrical energy from the PV panels (1) into alternating current (AC) electrical energy suitable for distribution,

[0060] • a power generation measurement system comprising: o a plurality of current sensors (6) configured to monitor the current flow within PV panels (1), and o a plurality of voltage sensors (7) configured to monitor the voltage across the PV panels (1);

[0061] • a weather monitoring station comprising: o a plurality of pyranometers (2) configured to measure irradiance, including global horizontal irradiance (GHI) and global inclined irradiance (GII), o a temperature sensor (3) configured to measure ambient temperature, o a humidity sensor (4) configured to measure ambient humidity, and o an anemometer (5) configured to measure wind speed.

[0062] Referring to Figure 2, the photovoltaic (PV) panels (1) comprise a plurality of PV modules (1) connected in series to form a string (10). Each string (10) may comprise up to thirty PV modules (1). Each string

[0063] (10) is operatively connected to an inverter (9) which converts DC (Direct Current) electrical energy from PV modules into AC (Alternative Current) electrical energy.

[0064] A plurality of strings (10), up to twenty-five or more, are connected to the input side of the inverter (9). In a central inverter (9) configuration, multiple strings (10) are connected to a DC combiner box, which combines the DC output of said strings (10) into a single DC output. The combined DC output is then fed into the DC side of the inverter (9).

[0065] Additionally, a plurality of sensors are deployed to measure various weather parameters and system performance metrics. The PV system is configured to efficiently convert incident radiation into electrical energy through the photovoltaic effect for utilization in a power grid

[0066] (11).

[0067] It is to be understood that the term “DC combiner box”, which is an important component in a photovoltaic (PV) power system. It functions as a central hub that consolidates the DC output from multiple strings of PV panels into a single, higher-current DC output. This consolidated DC output is subsequently fed into the inverter, which converts it into AC power suitable for distribution on the electrical grid.

[0068] While the aforesaid description provides context for the invention, it should be understood that the specific components and techniques described herein are well-known to those skilled in the art. This description is not intended to limit the scope of the invention.

[0069] For the purposes of this patent specification, the terms "photovoltaic module," "photovoltaic panel," and "PV generating station" are used interchangeably to refer to a system or collection of photovoltaic cells interconnected to generate electrical power from sunlight. While "photovoltaic module" and "photovoltaic panel" typically refer to a smaller, packaged unit of interconnected cells, and "PV generating station" refers to a larger-scale installation, in the context of this invention, these terms are used broadly to encompass any configuration of photovoltaic cells used for power generation, unless explicitly stated otherwise. Therefore, any reference to one of these terms should be understood to include the others, where applicable.

[0070] Now, according to the embodiment shown in Figure 3, a system and method for optimizing the performance of PV generating stations according to an exemplary embodiment are disclosed. The system comprises:

[0071] • a data collection module (14) configured to transmit said acquired sensors real-time data to a cloud platform through Modbus TCP / IP Protocol and Optical Fiber Communication (OFC) protocols,

[0072] • a data analysis module (15) configured to analyze said real-time data received from the data collection module (14) to predict PV power generation, optimize system performance, and identify potential issues, and • a user interface (16) configured to display said predictive PV power generation and system performance analysis received from the data analysis module (15) to provide accurate understandings and enable remote monitoring and control.

[0073] Further, according to the embodiment shown in Figure 4, said data collection module (14) is provided for monitoring and optimizing a photovoltaic (PV) generating station. The module comprises:

[0074] S14.1 : deploy plurality of sensors (2, 3, 4, 5) at the photovoltaic generating stations to acquire real-time data’s of irradiance, including global horizontal irradiance (GHI) and global inclined irradiance (GII), temperature, humidity and wind speed,

[0075] SI 4.2: transmission of the said sensors data to Supervisory Control and Data Acquisition (SCAD A) device through Modbus TCP / IP and Optical Fiber Communication (OFC) protocols, which ensures the storage of all sensor-related data to SCADA server,

[0076] SI 4.3: transmission of said sensors data from said SCADA system through OPC-UA (Open Platform Communication - Unified Architecture) to a cloud platform through computer networks.

[0077] Referring to the embodiment shown in Figure 5, said data analysis module (15) is provided for predicting future photovoltaic (PV) power generation. The module (15) uses advanced machine-learning techniques to analyse historical data and generate accurate predictions. Said data analysis module (15) consists of:

[0078] SI 5.1 : selection of various parameters such as current, voltage, temperature, radiation, wind speed, and humidity from the said sensor data stored on the cloud platform,

[0079] SI 5.2: transmission of said selected data for data preparation to filter out the anomalies, inconsistencies, missing values, generation forecasting, and extracting features such as weather-related and other PV parameters from the selected data, and generate a trained model using the filtered data through machine-learning technique such as Artificial Neural Networks and Random Forest; wherein, performance of said trained model datasets are tested using historical data through a hindcasting technique,

[0080] SI 5.3: detection of fault by tracking the actual generation and generation predicted by the said trained Al models for string-level performance and marking flag to the string having significant deviation from expected performance through advanced machine learning prediction techniques,,

[0081] SI 5.4: record rectification and correction of the said marked flag string and store said corrected data on the cloud platform storage for future analysis,

[0082] SI 5.5: calculation of the generation and revenue gain obtained by rectification of identified faulty strings.

[0083] Referring to Figure (6), said analysis module (15) generates the trained Al model using said filtered data through machine-learning techniques such as Artificial Neural Networks and Random Forest trains the model; wherein said training model consists of:

[0084] S15.21 : create a golden dataset using filtered data by applying a data preparation technique; wherein, said golden dataset accurately represents actual PV string performance under diverse environmental conditions;

[0085] SI 5.22: develop Al (Artificial Intelligence) models using advanced machine learning techniques, such as Artificial Neural Networks, Random Forests, and Support Vector Machines and train said Al models on 80% of said golden dataset to set the relationship between input features (e.g., weather conditions, system parameters) and output variables (e.g., PV power generation);

[0086] SI 5.23: evaluate said trained models on the remaining 20% of the golden dataset using metrics such as Mean Absolute Percentage Error (MAPE) to provide a conditional logic; wherein,

[0087] SI 5.231 : if said MAPE is below a predefined threshold, deploy the models for real-time PV power generation prediction;

[0088] SI 5.232: if said MAPE exceeds the threshold, fine-tune and retrain said models on the golden data; and iterate said fine-tuned and retrained recursively until satisfactory accuracy is achieved;

[0089] SI 5.24: deploy said trained and fine-tuned models to the cloudplatform, wherein said deployed models is used to predict PV power generation in real-time based on current and forecasted weather conditions, and system parameters and continuously monitor the performance of the deployed models and update the golden dataset with real-time data to ensure ongoing accuracy.

[0090] Said system performance evaluation thoroughly evaluates the system's performance using MAPE to assess the accuracy of power generation predictions and conduct comprehensive simulations to evaluate the model's predictive capabilities under various seasonal conditions.

[0091] It is to be understood that the term “golden dataset” used in the description refers to a golden dataset, which is mentioned here, is in the context of Machine Learning (ML) models which refers to a high- quality, labelled dataset that is used to train and evaluate the performance of these models. This dataset is sometimes referred to as the “ground truth,” and it is used to measure the performance of the Machine Learning Models. In accordance with the embodiment of the present invention, said user interface (16) is configured to display the overall performance of the system and feedback for user understanding. Further, in the event that the desired string performance is not achieved, said user interface (16) generates fault and delivers feedback to the user without human intervention. Subsequently, the system returns the data to the analysis module (15) for additional processing.

[0092] According to the exemplary embodiment shown in Figure 7, the method for generation prediction, fault detection, and performance enhancement of the photovoltaic generating station by the system is described herein. The method comprises of following steps: collecting the plurality of sensors (2, 3, 4, 5, 6, 7) data and transmitting said collected sensors real-time data to a cloud platform through a data collection module (14), analyzing said real-time data received from the data collection module (14) to predict PV power generation, optimize system performance, and identify potential issues through a data analysis module (15), and displaying said predicted PV power generation and system performance analysis received from the data analysis module (15) through a user interface module (16).

[0093] According to the embodiment shown in Figure 8, the method for said data-collecting module (14) consists of: deploying (S14.1), plurality of sensors (2, 3, 4, 5, 6, 7) at the photovoltaic generating stations to acquire real-time data’s of irradiance, including global horizontal irradiance (GHI), and global inclined irradiance (GII), temperature, humidity, string current, string voltage, and wind speed; and transmitting (SI 4.2), said sensors data to Supervisory Control and Data Acquisition (SCAD A) device through Modbus TCP / IP and Optical Fiber Communication (OFC) protocols, transmitting (SI 4.3), said sensors data from said SCADA system through OPC-UA (Open Platform Communication - Unified Architecture) to a cloud platform through computer networks, subsequently.

[0094] According to the embodiment shown in Figure 9, the method for said data analysis module (15) consists of: selecting (SI 5.1), various parameters from the said sensor data stored on the cloud platform, transmitting (S15.2), said selected data for data preparation to filter out the anomalies, inconsistencies, missing values, generation forecasting, and extracting features, and generating trained Al models using the filtered data through machine-learning techniques such as Artificial Neural Networks and Random Forest, detecting (SI 5.3), fault by tracking the actual generation and generation predicted by the said trained Al models for string-level performance and marking flag to the string having significant deviation from expected performance through advanced machine learning prediction techniques, correcting (SI 5.4), recording the rectification of said marked string, and storing said corrected data on the cloud platform storage for future analysis, calculating (SI 5.5), generation, and revenue gain obtained by rectification of identified faulty strings.

[0095] Now, referring to Figure 10, the method of model training involves the following steps: creating (S 15.21), a golden dataset using filtered data by applying a data preparation technique, developing (S15.22), Al (Artificial Intelligent) models using advanced machine learning techniques, and training said Al models on 80% of said golden dataset to set the relationship between input features and output variables; evaluating (SI 5.23), said trained models on the remaining 20% of the golden dataset using metrics such as Mean Absolute Percentage Error (MAPE); wherein, o if said MAPE is below a predefined threshold (S15.231), deploy the models for real-time PV power generation prediction; o if said MAPE exceeds the threshold (S 15.232), fine-tune and retrain said models on the golden data; and iterate said fine-tuned and retrained recursively until satisfactory accuracy is achieved; deploying (S15.24), said trained and fine-tuned models to the cloudplatform.

[0096] For purposes of clarity, it is to be understood that the terms "actual data" and "real-time data" are used interchangeably herein to refer to the same data stream. These terms include the measured values collected from the sensors at the PV generating station, reflecting the present state of the environment and PV system performance.

[0097] The system and method for predicting photovoltaic (PV) power generation, detecting faults, and enhancing the performance of PV generating stations are disclosed. The system utilizes Supervisory Control and Data Acquisition (SCADA) data collected from sensors installed at PV generating stations to identify anomalies in the station’s performance and measurements. The system improves the accuracy of PV power generation predictions, enhances fault detection capabilities, reduces the occurrence of false alarms and alerts, and optimizes the performance and maintenance of PV generating stations.

[0098] Referring to the embodiment described in Figure 11, the user interface (16) provides comprehensive information regarding key performance indicators at both portfolio and station levels. It displays the number of issues identified, closed, and open for each generating station, with an aggregated summary for the entire portfolio. Additionally, the user interface categorizes issues and presents the associated generation and revenue gains realized through issue resolution.

[0099] The user interface (16) empowers users to navigate through various issue tickets generated for the plant and update the corresponding corrective actions. Furthermore, users may effortlessly visualize expected generation, actual generation, and irradiation values for any specific string within the generating station through a graphical representation .

[0100] In the foregoing, it is to be understood that the terms “Modbus TCP / IP”, “OFC”, “SCADA system”, “OPC server”, “anomaly detection”, and “Al module” are intended to have their broadest meaning as understood by those skilled in the art.

[0101] The present invention has been described in terms of specific embodiments, and it is understood that various other modifications may be made without departing from the scope of the invention. The above-described embodiments are illustrative and not restrictive. The application is thus intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims. Experimental results

[0102] The embodiments described herein above have undergone extensive experimentation to determine the system's corrective capabilities in scenarios where string performance is compromised due to faulty or damaged modules, in comparison to strings with healthy modules. The desired outcome should be a substantial degree of correction for faulty or damaged modules. The graphical representation of said test results is shown in Figures 12 to 15.

[0103] Case 1 : As depicted in Figure 12, on day 1 , a significant discrepancy was observed between predicted and actual generation, resulting in the generation of a ticket. Upon inspection, the root cause of the deviation was identified as soiling. The photovoltaic (PV) panels were subsequently cleaned on day 2, leading to a noticeable improvement in performance as evidenced by the results on day 3.

[0104] Case 2: Now referring to the figure shown in Figure 13, a string breakdown occurred on day 1, resulting in a substantial gap between predicted and actual generation and the subsequent generation of a ticket. In response to the ticket, corrective actions were implemented, and the string was successfully restored on day 2.

[0105] Case 3: Figure 14 illustrates, that on day 7, a ticket was generated due to a significant gap between actual and expected generation. Upon inspection, a tracker movement-related issue was identified and promptly rectified, resulting in improved performance as observed on day 2.

[0106] Case 4: As shown in Figure 15, the performance of a string was compromised due to three damaged modules. This performance degradation led to the generation of a ticket, and subsequent inspection revealed the presence of damaged modules. After replacing the damaged modules, a noticeable performance improvement was observed.

[0107] ADVANTAGES OF THE PRESENT INVENTION:

[0108] A system and method for generation prediction, fault detection, and performance enhancement of photovoltaic generating station as illustrated in the above description of the invention provides the following advantages:

[0109] • that eliminates redundant data points from historical station performance.

[0110] • that ensures that the golden dataset is continuously updated to incorporate changes in generating system configuration and parameters.

[0111] • that is easily implemented across plants with different configurations or technologies.

[0112] • that establishes benchmark performance and identifies performance issues.

[0113] • that notifies the site managers about the detected issues.

[0114] • that updates corrective action taken and closes the raised tickets.

[0115] • that has a user-friendly, intuitive graphical user interface on which various performance KPIs and savings can be visualized, enabling efficient plant monitoring and issue rectification.

[0116] • that enables early identification of the smallest performance issues in a PV generating station, resulting in significant cost savings and increasing plant longevity.

[0117] • that minimizes generation loss of green energy and is thus environmentally friendly.

Claims

AMENDED CLAIMS received by the International Bureau on 15 October 2025 (15.10.2025)We claim,1. A system and method for photovoltaic generating station; wherein said system comprises:— a data collection module (14) configured to transmit said acquired sensors real-time data to a cloud platform through Modbus TCP / IP Protocol and Optical Fiber Communication (OFC) protocols, and— a user interface (16) configured to display said predicted string level PV power generation and system performance analysis received from the data analysis module (15) and provide accurate understandings and enable remote monitoring and control, characterized in that, the system further comprises:— a data analysis module (15) configured to analyze system performance by said real-time data received from the data collection module (14) and predict PV power generation, and identify potential issues, wherein said data analysis module (15) is configured for: transmission (S15.2) of said selected data for data preparation to filter out the anomalies, inconsistencies, missing values, generation forecasting, and extracting features such as weather-related and other PV parameters from the selected data, and generate a trained Al models using the filtered data through machine-learning technique such as Artificial Neural Networks and Random Forest; wherein, performance of said trained model datasets are tested using historical data through a hindcasting technique,detection (S15.3) of fault by tracking the actual generation and generation predicted by the said trained Al models for string-level performance and mark a flag to the string having significant deviation from expected performance through advanced machine learning prediction techniques,2. The system as claimed in claim 1, wherein said data collection module (14) consists of:— deploy (S14.1) plurality of sensors (2, 3, 4, 5, 6, 7) at the photovoltaic generating stations to acquire real-time data’s of irradiance, including global horizontal irradiance (GHI) and global inclined irradiance (GII), temperature, humidity, current, voltage and wind speed,— transmission (S14.2) of the said sensors data to Supervisory Control and Data Acquisition (SCADA) device through Modbus TCP / IP and Optical Fiber Communication (OFC) protocols,— transmission (S14.3) of said sensors data from said SCADA system through OPC-UA (Open Platform Communication - Unified Architecture) to a cloud platform through computer networks.

3. The system as claimed in claim 1, wherein said data analysis module (15) further consists of:— selection (SI 5.1) of various parameters such as current, voltage, temperature, radiation, wind speed, and humidity from the said sensor data stored on the cloud platform,— record (S15.4) rectification and correction of the said marked flag string and store said corrected data on the cloud platform storage for future analysis, and— calculation (S15.5) of the generation and revenue gain obtained by rectification of identified faulty strings.

4. The system as claimed in claim 1, wherein said analysis module (15) generates the trained Al models using said filtered data consists of:— create (S15.21) a golden dataset using filtered data by applying a data preparation technique; wherein, said golden dataset accurately represents actual PV string performance under diverse environmental conditions;— develop Al (Artificial Intelligence) (S 15.22) models using advanced machine learning techniques, such as Artificial Neural Networks, Random Forests, and Support Vector Machines and train said Al models on 80% of said golden dataset to set the relationship between input features (e.g., weather conditions, system parameters) and output variables (e.g., PV power generation);— evaluate (S 15.23) said trained models on the remaining 20% of the golden dataset using metrics such as Mean Absolute Percentage Error (MAPE) to provide a conditional logic; wherein, o if said MAPE (S 15.231) is below a predefined threshold, deploy the models for real-time PV power generation prediction; o if said MAPE (S 15.232) exceeds the threshold, fine-tune and retrain said models on the golden data; and iterate said fine-tuned and retrained recursively until satisfactory accuracy is achieved;— deploy (S 15.24) said trained and fine-tuned models to the cloud-platform, wherein said deployed models is used to predict PV power generation in real-time based on current weather conditions, and system parameters and continuously monitor the performance of the deployed models and update the golden dataset with real-time data to ensure ongoing accuracy.

5. A method for photovoltaic generating station system as claimed in claim 1 comprises of following steps:— collecting the plurality of sensors (2, 3, 4, 5, 6, 7) data and transmitting said collected sensors real-time data to a cloud platform through a data collection module (14), and— displaying said predicted PV power generation and system performance analysis received from the data analysis module (15) through a user interface module (16) characterized in that, the method further comprises of following steps:— analyzing system performance by said real-time data received from the data collection module (14) and predicting PV power generation, and identifying potential issues through a data analysis module (15), wherein, said analyzing step is being performed through the data analysis module (15) by: transmitting (S15.2), said selected data for data preparation to filter out the anomalies, inconsistencies, missing values, generation forecasting, and extracting features, and generating trained model datasets of the filtered data through machine-learning techniques such as Artificial Neural Networks and Random Forest,detecting (S15.3), fault by tracking the actual generation and generation predicted by the said trained Al models for string-level performance and marking flag to the string having significant deviation from expected performance through advanced machine learning prediction techniques.

6. The method as claimed in claim 5, wherein said step of collecting the plurality of sensors (2, 3, 4, 5, 6, 7) data by the data collection module (14) consists of the following steps:— deploying (S14.1), plurality of sensors (2, 3, 4, 5, 6, 7) at the photovoltaic generating stations to acquire real-time data’s of irradiance, including global horizontal irradiance (GHI), and global inclined irradiance (GII), temperature, humidity, current, temperature, and wind speed; and— transmitting (S14.2), said sensors data to Supervisory Control and Data Acquisition (SCADA) device through Modbus TCP / IP and Optical Fiber Communication (OFC) protocols,— transmitting (S14.3), said sensors data from said SCADA system through OPC-UA (Open Platform Communication - Unified Architecture) to a cloud platform through computer networks, subsequently.

7. The method as claimed in claim 5, wherein said step of analyzing system performance by the data analysis module (15) consists of the following steps:— selecting (S15.1), various parameters from the said sensor data stored on the cloud platform,— recording the rectification (S15.4), and correcting of said marked string, and storing said corrected data on the cloud platform storage for future analysis,— calculating (S15.5), generation, and revenue gain obtained by rectification of identified faulty strings.

8. The method as claimed in claim 5, wherein said analysis module (15) generates the trained Al models using said filtered data consists of the following steps:— creating (S15.21), a golden dataset using filtered data by applying a data preparation technique,— developing (S 15.22), Al (Artificial Intelligent) models using advanced machine learning techniques, and training said Al models on 80% of said golden dataset to set the relationship between input features and output variables;— evaluating (S 15.23), said trained models on the remaining 20% of the golden dataset using metrics such as Mean Absolute Percentage Error (MAPE); wherein,— if said MAPE is below a predefined threshold (S15.231), deploy the models for real-time PV power generation prediction;— if said MAPE exceeds the threshold (S 15.232), fine-tune and retrain said models on the golden data; and iterate said finetuned and retrained recursively until satisfactory accuracy is achieved;— deploying (S 15.24), said trained and fine-tuned models to the cloud-platform.[0001][0002]STATEMENT UNDER ARTICLE 19 (1)[0003]With respect to the statement reasoned under Box no. VII of the PCT search report with a written opinion dated 18 / 07 / 2025, we amend the claims under Article 19 to meet with the criteria of Article 6 PCT.[0004]The amendments include amending the independent claims (claim 1 and claim 5) in two-part form that clearly delineate the features already known from the prior art (placed in the preamble) from the novel and inventive features of the present invention (placed in the characterizing part). These amendments do not add new matter and are submitted to bring the claims into conformity with the formal requirements. These amendments are made to clarify the invention in response to the objections raised by the Patent Examiner.[0005]Fresh sheets of marked copy of amended claims and clean copy of amended claims are submitted and enclosed herewith as Annexures 1 and 2, respectively, which please take on record.

Images