New energy intelligent power station inspection analysis and monitoring system adaptive to few-man on duty

Abstract

The application discloses a new energy intelligent power station inspection analysis and monitoring system suitable for few-man on duty, relates to the technical field of power station operation and maintenance monitoring, and comprises the following: a field sensing and collecting module, which is used for arranging multiple types of field sensing and collecting terminals; a network transmission module, which is used for constructing a station-in-fusion communication transmission link; an edge computing processing module, which is arranged on a power station edge gateway device and is used for performing protocol analysis, data cleaning, noise filtering and format standardization processing on front-end heterogeneous collected data; a cloud intelligent analysis module, which is used for receiving standardized data sets output by the edge computing processing module; a comprehensive application service module, which is used for realizing visual presentation of power station global operation data, receiving cloud fault research and judgment results and performing multi-channel graded alarm pushing; generating operation and maintenance priorities according to fault emergency degree weights, automatically generating targeted operation and maintenance work orders and directing dispatching; and supporting operation and maintenance work order disposal, acceptance and archiving whole-process closed-loop management.

Description

Technical Field

[0001] This invention relates to the field of power plant operation and maintenance monitoring technology, and in particular to a smart new energy power plant inspection, analysis and monitoring system adapted for minimal staffing. Background Technology

[0002] The photovoltaic power generation industry has entered a stage of rapid large-scale development, with a large number of centralized and distributed photovoltaic power plants being put into operation simultaneously in the region. The installed capacity of power plants continues to expand. The traditional on-site, full-staff operation and maintenance model is characterized by high labor costs and low management efficiency. As a result, the industry is gradually promoting new operation and maintenance solutions with fewer on-site personnel and remote centralized control. Existing photovoltaic power plant monitoring and operation and maintenance systems on the market are generally equipped with basic data acquisition modules and simple back-end display platforms. They can collect, store, and display basic visualizations of operating parameters of core power generation equipment such as photovoltaic strings, inverters, and transformer substations. At the same time, they can output simple fault alerts based on preset parameter thresholds, which can replace manual on-site meter reading to a certain extent.

[0003] However, existing operation and maintenance monitoring solutions still have significant shortcomings. The core technical deficiencies are: the entire system lacks a complete digital operation and maintenance closed-loop management system; fault early warning, operation and maintenance task allocation, on-site fault handling, and completion acceptance archiving are all fragmented, and data cannot be automatically transferred between different business processes. After receiving fault alerts, operation and maintenance personnel need to manually sort out the fault level and location, develop repair plans offline, and register repair records. All operation and maintenance work heavily relies on manual coordination and paper-based records, which not only increases a large amount of repetitive manual work but also easily leads to problems such as chaotic fault handling sequences, lack of record-keeping during operation and maintenance, and inability to trace fault information. Summary of the Invention

[0004] To address the aforementioned issues, this invention provides a smart power plant inspection, analysis, and monitoring system adapted for minimal human intervention.

[0005] The present invention provides an inspection, analysis, and monitoring system for new energy smart power plants with minimal human intervention, comprising: The field sensing and acquisition module is used to deploy various types of field sensing and acquisition terminals to continuously collect raw operational data from all dimensions of the power plant's power generation equipment, power distribution equipment, and the station environment, achieving blind-spot-free sensing and acquisition of all station equipment. The network transmission module is used to build an integrated communication transmission link within the station, carrying bidirectional transparent transmission and unified access of front-end sensing and collection data; it monitors the operation status of the communication link in real time, identifies and marks link anomalies, and ensures stable and complete data transmission across levels. The edge computing processing module is deployed on the power plant edge gateway device. It is used to perform protocol parsing, data cleaning, noise filtering, and format standardization on the heterogeneous data collected from the front end. It also removes invalid and interference data through a steady-state judgment algorithm. It is configured with local data caching and breakpoint resume mechanism to complete the preliminary judgment of simple on-site working conditions. The cloud-based intelligent analysis module receives standardized datasets output by the edge computing processing module, performs quantitative calculations based on a preset equipment operating condition analysis model, automatically identifies various explicit and implicit abnormal operating conditions of the equipment, and completes fault classification, categorization, and source tracing. The integrated application service module is used to realize the visualization of power plant operation data, receive cloud-based fault assessment results and execute multi-channel hierarchical alarm push; generate maintenance priorities based on fault urgency weights, automatically generate targeted maintenance work orders and distribute them in a targeted manner; support closed-loop management of the entire process of maintenance work order handling, acceptance and archiving, and iterate and optimize the cloud-based analysis and judgment model by feeding back maintenance business data.

[0006] In summary, the present invention has at least the following beneficial effects: 1. This invention relies on automatic on-site data collection, edge data purification, and multi-dimensional algorithm joint analysis in the cloud, eliminating the need for maintenance personnel to be stationed on-site regularly and conduct full-domain inspections, thus significantly reducing on-site manpower input; at the same time, it automatically identifies explicit faults and hidden hazards such as power attenuation, solving the problems of missed detection and lag in manual inspections, and reducing manual maintenance costs from the root.

[0007] 2. This invention automatically generates dedicated maintenance work orders based on a quantitative scoring system for fault urgency, dispatching maintenance tasks only to abnormal locations. This eliminates the traditional blind inspection model that covers all areas, reducing the ineffective consumption of personnel, vehicles, and consumables, and significantly improving the efficiency of maintenance resource utilization. At the same time, it establishes an integrated maintenance closed loop of "alarm-dispatch-handling-archiving-data feedback". All maintenance records are stored online and traceable. On-site work order handling data is fed back into the cloud analysis model to continuously correct fault judgment thresholds and classification weights, thereby continuously improving the accuracy of fault identification and reducing false alarms and missed alarms.

[0008] 3. This invention completes the standardized processing of multi-protocol device data through edge gateways, breaks down data silos between various devices and subsystems, and realizes centralized and distributed power station linkage monitoring within a region based on a unified visualization platform, without the need to switch between multiple systems, thus greatly improving the efficiency of centralized management of large-scale power stations.

[0009] 4. The edge layer of this invention performs data noise reduction and cache breakpoint resume on-site, filters on-site interference data, and reduces the computing power pressure on the cloud. Combined with the cloud-based operating condition deviation and trend analysis model, it not only ensures data integrity and no loss in network outage scenarios, but also improves the accuracy of equipment anomaly identification, realizing 24-hour uninterrupted online intelligent monitoring of the power station.

[0010] 5. This invention removes interference data through edge data preprocessing and combines it with a cloud-based multi-dimensional quantitative analysis model to simultaneously identify explicit faults and slow-decaying hidden dangers. It can provide early warning of faults and accurately locate abnormal points, changing the problems of delayed fault detection and long-term operation of equipment with defects caused by traditional manual inspections. It shortens the fault handling window, reduces power generation losses caused by equipment abnormalities, extends the service life of photovoltaic and power distribution equipment, and reduces equipment maintenance and replacement costs. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic diagram of the overall system architecture of the present invention; Figure 2 This is a schematic diagram of the field sensing and acquisition module architecture of the present invention; Figure 3 This is a diagram of the fault alarm analysis interface of the present invention; Figure 4 This is a diagram of the operation and maintenance work order dispatch interface of the present invention. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] The following is in conjunction with the appendix Figures 1 to 4 The present invention will be described in further detail below.

[0015] like Figure 1 As shown, the inspection, analysis and monitoring system for new energy smart power plants with minimal human intervention includes a field perception and acquisition module, a network transmission module, an edge computing processing module, a cloud-based intelligent analysis module and a comprehensive application service module.

[0016] like Figure 2As shown, the field perception and acquisition module, as the lowest-level global data perception unit of this system, is the foundation for the original data supply of the entire system. It realizes synchronous acquisition of all equipment, all measurement points, and all operating conditions of centralized and distributed new energy photovoltaic power stations without blind spots, completely replacing the traditional manual field meter reading, manual visual inspection, and manual equipment status recording. It provides continuous, real-time, complete, and quantitative original operating datasets for upper-level network transmission, edge preprocessing, and cloud-based intelligent analysis.

[0017] This module is composed of multiple standardized sensing and acquisition terminal hardware components, including five types of acquisition hardware: photovoltaic string acquisition terminals, inverter acquisition modules, transformer substation monitoring and control devices, environmental monitoring sensors, and on-site video acquisition equipment. Each type of terminal is deployed in layers and areas according to the power station equipment layout, covering all key locations in the photovoltaic array area, inverter room, transformer substation distribution room, and outdoor public areas of the station. Each acquisition terminal independently completes high-frequency sampling of data at its corresponding measurement point and synchronously outputs standardized raw acquisition signals.

[0018] The photovoltaic string acquisition terminal is connected one-to-one to the output end of each photovoltaic string. It uses a high-frequency polling sampling mechanism to continuously collect the core quantitative parameters of the single photovoltaic string. The collected indicators include the string's real-time operating voltage, real-time output current, instantaneous output power, terminal operating temperature, branch load rate, and device communication heartbeat status. The sampled data is accompanied by the acquisition terminal's unique device number and a millisecond-level high-precision timestamp for synchronous output.

[0019] The real-time output power of a single device is generated in real time by the formula P=U×I, where P is the real-time output power of the string (W), U is the real-time operating voltage of the string (V), and I is the real-time operating current of the string (A). Time-series power data is obtained by continuous sampling and used for back-end power deviation analysis, string attenuation identification, and power generation loss calculation.

[0020] The inverter acquisition module connects to each photovoltaic inverter in the station through a standard industrial communication interface to continuously collect the inverter's overall operating parameters. The collected data includes the inverter's AC / DC side voltage, AC / DC output current, real-time output power, IGBT module operating temperature, cooling fan operating status, overall load conditions, inverter internal fault alarm native signals, and equipment communication heartbeat messages. The module also records the inverter's start-up and shutdown times, grid connection status, and islanding protection action records, thus fully capturing the inverter's all-dimensional operating status on the generator side.

[0021] The monitoring and control device for the prefabricated transformer is deployed inside each prefabricated transformer, focusing on the full coverage acquisition of the operating conditions on the distribution side. The core acquisition includes the voltage of the three-phase power grid AB line, the voltage of the BC line, the voltage of the CA line, the transformer winding temperature, the load current on the high and low voltage sides, the transformer oil temperature, the tap changer position, the circuit breaker opening and closing status, and the communication heartbeat status of the distribution circuit. The device captures in real time the original characteristic data of distribution anomalies such as voltage deviation, three-phase imbalance, and overload temperature rise on the distribution side, supporting cloud-based distribution fault identification and power quality analysis.

[0022] Environmental monitoring sensors are deployed in unobstructed locations such as open areas of the photovoltaic array, rooftops of the power station, and outdoor platforms of transformer substations to collect basic environmental parameters of the power station in real time, including solar irradiance, ambient air temperature and humidity, wind speed, and rainfall. The collected environmental time-series data will be linked to the photovoltaic strings and inverter power generation data one by one, serving as a benchmark for cloud-based quantitative calculation of operating condition deviations. This helps to distinguish between environmental fluctuations and equipment malfunctions, reducing the probability of false alarms in the system.

[0023] On-site video acquisition equipment is deployed in key areas such as the main road of the photovoltaic array, the entrance and exit of the inverter room, the transformer substation, and the main gate of the site, according to the regional coverage requirements. It continuously collects real-time visual images and metadata of the site equipment. The video metadata includes real-time video streams, real-time images of the equipment area, image acquisition timestamps, and geographical identifiers of the equipment locations. On the one hand, it allows the integrated application service layer to remotely visualize the real-time scene. On the other hand, it works with cloud-based image recognition algorithms to help identify visually impaired faults such as equipment damage, cable detachment, and obstruction by foreign objects.

[0024] The network transmission module is located between the field sensing and acquisition module and the edge computing processing module. As a unified transmission channel for cross-level data communication in the system, it establishes a wired and wireless converged communication link for photovoltaic power plants. It carries the bidirectional transparent transmission and unified access of all original messages from sensing and acquisition terminals. It has functions such as multi-protocol forwarding and scheduling, link redundancy backup, transmission security encryption, full-link status monitoring, and anomaly marking. It can identify link anomalies such as disconnection, high latency, high packet loss, wireless signal degradation, and traffic overload, ensuring stable and complete data transmission. It overcomes the shortcomings of traditional power plants, such as scattered links, incompatible protocols, lack of fault monitoring, and interruption of data acquisition due to single-point link failure.

[0025] The network transmission module includes a multi-protocol forwarding and scheduling unit, a link backup and switching unit, a transmission security encryption unit, and a link monitoring and processing unit. The link monitoring and processing unit further comprises a heartbeat message sending and receiving subunit, a data packet statistics and verification subunit, a link status storage subunit, and an anomaly identification encapsulation subunit. The multi-protocol forwarding and scheduling unit is compatible with mainstream field communication protocols such as Modbus, LoRa, MQTT, RTSP, and IEC60870, stripping various terminal message protocol encapsulations and extracting raw measurement point data. The unit divides the data into three service buffer queues: real-time acquisition, video, and heartbeat, and configures differentiated bandwidth priorities to ensure priority transmission of equipment operating data. It supports the fragmentation and reassembly of ultra-large messages at the receiving end, while also distinguishing between uplink acquisition data and downlink control commands to achieve bidirectional scheduling and forwarding.

[0026] The link backup switching unit configures primary and backup dual links for the acquisition terminal. Under normal circumstances, the primary link carries the business and the backup link is only kept alive by heartbeat. The unit reads the link monitoring status flags. When the primary link is abnormal, it switches to the backup link for transmission in milliseconds. After the primary link indicators recover and stabilize, it smoothly switches back to traffic. It synchronously records the link switching time, link type, data loss and generates logs to push to the edge computing processing module.

[0027] The transmission security encryption unit encrypts and protects all station service messages, heartbeat messages, and remote control commands. Uplink measurement point data is encrypted using AES-128, and each site is configured with an independent isolation key. Downlink control commands verify the device's exclusive digital signature, and illegally signed messages are discarded directly. Hash verification is added on top of CRC verification to achieve double anti-tampering, and only data frames that pass the verification can be forwarded upward.

[0028] The link monitoring and processing unit relies on the collaboration of its internal sub-units to complete the full life cycle operation monitoring and link anomaly marking of the communication link. The heartbeat message sending and receiving sub-unit pre-configures a unified heartbeat sending cycle for all communication links within the station, covering various transmission links such as wired Ethernet links, fiber optic communication links, 4G / 5G public wireless links, and LoRa local wireless links. For each independent communication link, it periodically generates a heartbeat detection message carrying a unique link identifier code and sends it to the corresponding sensing and acquisition terminal at the end of the link.

[0029] After receiving a heartbeat detection message, the end-point sensing and acquisition terminal immediately sends back a response message carrying the current link identifier and a high-precision timestamp of the local message reception. The heartbeat message sending and receiving subunit synchronously records the time of each detection message being sent and the time of the response message being received, automatically calculates the round-trip transmission delay of a single message, and continuously counts the number of heartbeat detections that have not received a valid response message within a preset monitoring period. The statistical results are used as the core quantitative indicator for determining link connectivity.

[0030] The data packet statistics and verification subunit adds continuously increasing frame numbers to service data frames transmitted through each communication link, performs CRC cyclic redundancy check on each frame of service data, and determines the integrity of single-frame data transmission based on the verification results. The subunit counts the total number of data frames sent by the edge gateway and the number of valid data frames returned by the terminal that have passed verification within a unit of time for each link. It calculates the real-time packet loss rate of the link by combining the total number of frames and the number of valid received frames. It also collects the RSSI signal strength of the wireless link and the real-time bandwidth occupancy of the wired link port. All monitoring indicators, including link round-trip delay, periodic packet loss rate, wireless signal strength, and port bandwidth occupancy, are output to the link status storage subunit in real time.

[0031] The link state storage subunit builds a standardized link state data table in the gateway's local non-volatile memory medium. The data table is pre-set with multiple fields such as link number, link type, real-time round-trip transmission delay, periodic packet loss rate, number of consecutive heartbeat no-response, wireless signal strength, link normal / abnormal status flag, link primary / backup attributes, and link switching history. The link state storage subunit receives all link monitoring indicators output by the heartbeat message sending and receiving subunit and the data packet statistics and verification subunit in real time, and writes them into the link state data table one by one to complete local persistent caching. This data can be retrieved and compared in real time by the anomaly identification encapsulation subunit and the link backup switching unit.

[0032] The anomaly identification encapsulation subunit has a pre-stored set of differentiated link anomaly judgment thresholds. The threshold set includes the link delay exceeding the limit threshold, the packet loss rate exceeding the limit threshold, the number of consecutive heartbeat no response threshold, the wireless signal strength lower limit threshold, and the port bandwidth overload threshold. The anomaly identification encapsulation subunit reads the real-time operation indicators of each link in the link status data table in real time and compares the measured indicators with the corresponding thresholds item by item.

[0033] When any measured indicator exceeds the preset threshold range, it is determined that there is a transmission anomaly in the current link. A link anomaly label is automatically generated, and the anomaly label, the value of the indicator exceeding the limit, and the high-precision timestamp of the anomaly trigger are uniformly encapsulated into the additional field of the header of all service data frames currently transmitted on this link, thus completing the pre-marking of the link anomaly status. For normal links where all indicators are within the threshold range, a link normal identification field is added to the header of the service data frame and transmitted synchronously to the upstream edge computing processing module along with the collected data.

[0034] The edge computing processing module is deployed within the power plant's edge gateway, positioned between the network transmission module and the cloud-based intelligent analysis module. It performs on-site preprocessing of heterogeneous front-end collected data and outputs standardized datasets for uploading to the cloud. It also features local monitoring during network outages, remote command relay, local alarms, and log retention capabilities, reducing cloud computing power consumption, ensuring data integrity and basic anomaly alerts in offline scenarios, and adapting to the 24 / 7 data flow requirements of minimally staffed power plants. The module receives raw packets with link markers and encrypted verification fields from the network transmission module, sequentially performing packet parsing and verification, interference data filtering, deduplication and normalization, local caching, and initial local condition assessment. When connected to the network, it pushes standardized data to the cloud; when offline, it stores data locally and triggers local alarms, while simultaneously relaying device control commands from the cloud.

[0035] The edge gateway is equipped with a data preprocessing unit, an edge local control forwarding unit, an edge local alarm unit, and an edge log storage unit. The data preprocessing unit includes a data protocol parsing subunit, a data security verification subunit, a data cleaning subunit, a data deduplication subunit, a format normalization subunit, a local storage subunit, and a simple working condition discrimination subunit.

[0036] The data protocol parsing subunit is compatible with field communication protocols such as Modbus, LoRa, MQTT, RTSP, and IEC60870. It identifies the message protocol identifier, strips the encapsulation header, extracts valid service data such as device number, timestamp, measurement point value, and link status, and outputs the raw measurement point data to the downstream.

[0037] The data security verification subunit performs CRC verification, AES decryption, and device digital signature verification in sequence. If any verification fails, the message is discarded directly, and only legitimate data enters the data cleaning subunit.

[0038] The data cleaning subunit receives legitimate real-time power, voltage, current, and temperature sampling values ​​from the data security verification subunit. The subunit uses a fixed-length sliding window FIFO memory queue to continuously cache consecutive time-series sampling data. When the storage space of the sliding window reaches a preset capacity threshold for newly added sampling data, the earliest time-series historical sampling record stored at the front of the queue is automatically removed. The data cleaning subunit reads all valid power-related sampling values ​​within the sliding window in real time and calculates the short-time average power corresponding to all sampling values ​​within the window using an arithmetic mean. The subunit's local non-volatile storage medium pre-stores power fluctuation thresholds categorized by device type and installed capacity, targeting three core types: photovoltaic strings, inverters, and transformer substations. Each device is configured with an independent fluctuation threshold parameter. The data cleaning subunit calculates the absolute difference between a single real-time power sample value and the short-term average power line by line. The calculated difference is compared with the preset allowable fluctuation threshold matched by the current device. If the absolute difference between the real-time power and the short-term average power is greater than the preset allowable fluctuation threshold, the current sampled data is determined to be invalid interference data caused by instantaneous fluctuations in station illumination, electromagnetic interference, or instantaneous sampling jumps at the terminal. The complete sampled message is discarded and not forwarded. If the difference is less than or equal to the preset allowable fluctuation threshold, the sampled data is determined to be valid steady-state data that can reflect the true operating state of the device. The complete sampled data is retained and transmitted to the downstream data deduplication subunit.

[0039] The data deduplication subunit receives the steady-state valid sampled data output by the data cleaning subunit. The subunit selects the device's unique hardware number and millisecond-level high-precision acquisition timestamp as a joint uniqueness verification identifier. It traverses the current data buffer queue and performs matching and retrieval comparison on multiple identically sourced acquisition messages reported repeatedly under the same device and the same millisecond acquisition timestamp. When a duplicate identifier message is found, only the first original sampled data message that arrives earliest in the time sequence is retained, and all other duplicate reports with the same identifier are automatically deleted. This completely eliminates the data redundancy accumulation problem caused by multi-link concurrent reporting, terminal disconnection retransmission, and gateway polling duplicate acquisition, effectively reducing the total number of uplink transmission messages and saving bandwidth resources for the site's wireless network and fiber optic transmission.

[0040] The format normalization subunit receives the non-repeating steady-state measurement point data output by the data deduplication subunit. The subunit has a built-in unified standardized data structure template. The template uniformly specifies the field length, data encoding format, numerical precision retention digits, and unit identifier for the equipment number field, acquisition timestamp field, equipment type field, three-phase voltage measurement point field, current measurement point field, power measurement point field, temperature measurement point field, environmental irradiance temperature and humidity field, link status mark field, and local anomaly mark field. The subunit maps and fills the corresponding fields of the standardized structure with the deduplicated values ​​of various heterogeneous measurement points one by one, uniformly converts floating-point, integer, and string data formats, and automatically fills the null value identifier field for missing measurement points. This completes the unified and standardized data format of all types of equipment. The standardized measurement point data with completely consistent output field structure is transferred to the local storage subunit, eliminating the cloud parsing and adaptation pressure caused by inconsistent data fields and precision of data reported by different terminals.

[0041] The local storage subunit divides the edge gateway's local non-volatile storage medium into three independent and physically isolated storage blocks: a real-time data cache partition, an offline backup cache partition, and an edge operation log partition. The real-time data cache partition is used to temporarily store real-time standardized sampling data when the communication link is normally connected. The offline backup cache partition specifically stores all standardized raw data continuously generated at the site during offline periods when the communication link is disconnected. The edge operation log partition independently stores operation events, packet drop records, link switching records, local alarm records, and breakpoint resume operation records for each unit within the edge gateway. The local storage subunit uniformly binds a time-series acquisition timestamp, a unique identifier for the acquisition device, and a globally generated data sequence number to each standardized measurement point data stored in the cache partition, synchronously recording the precise time the data is written to the local cache. The system incorporates a built-in breakpoint resume scheduling logic. This logic continuously reads the link connectivity status flag field encapsulated in the data frame header by the network transmission module. When the link status flag indicates a disconnection, it automatically pauses the push of real-time standardized data to the cloud-based intelligent analysis module and continuously writes all newly generated standardized measurement point data into the offline backup cache partition for complete retention. When the link status flag returns to normal connectivity, the breakpoint resume scheduling logic iterates through and reads all offline cached data within the offline backup cache partition. Based on the time-series acquisition timestamp bound to each data entry, it performs a global ascending sort and resends the offline cached data to the cloud-based intelligent analysis module in batches and segments according to the chronological order. After all offline cached data has been completely resent, the system automatically switches back to the real-time synchronous push mode, ensuring that all collected data during the offline network period is collected without loss and can be fully traced back.

[0042] The simplified operating condition discrimination subunit pre-stores a set of basic threshold parameters categorized by equipment type. These parameters are configured according to four main categories: photovoltaic strings, inverters, transformer substations, and environmental monitoring sensors. The subunit sets corresponding upper and lower limits for voltage, current, output power, equipment winding / terminal temperature, and normal irradiation range for each measurement point. The simplified operating condition discrimination subunit reads the standardized measurement point data (after cleaning, deduplication, and format normalization) line by line, extracts the real-time values ​​of each measurement point within the data structure, and matches them with the basic threshold parameters for the current equipment type. Numerical comparison calculations are performed item by item. When the real-time value of the measuring point exceeds the upper and lower limits of the corresponding threshold, an on-site anomaly screening mark field is added to this standardized data structure, and the name of the measuring point exceeding the limit, the difference exceeding the limit, and the duration of the exceeding limit are simultaneously marked. When all the values ​​of the measuring point are within the normal threshold range, an equipment normal operation mark field is added to this standardized data. The standardized dataset carrying the on-site screening mark and complete measuring point information is uniformly output to the cloud intelligent analysis module, providing a preliminary on-site screening reference for multi-dimensional in-depth fault analysis and fault classification and tracing in the cloud.

[0043] The edge local control forwarding unit receives remote control and parameter configuration instructions from the cloud. After completing the signature and permission verification, it adapts to the corresponding terminal communication protocol and sends them to the field acquisition terminal via the network transmission module. Simultaneously, it records the instruction issuance and execution logs and stores them in the edge log storage unit.

[0044] The edge local alarm unit reads data with local anomaly tags, classifies local alarms into two levels based on the extent of exceeding limits, pushes anomaly information through gateway audio-visual devices and local operation and maintenance mini-programs, and archives alarm records synchronously for storage, which are then uploaded to the cloud after connecting to the network.

[0045] The edge log storage unit uniformly collects various operational logs such as message processing, link switching, breakpoint resumption, local alarms, remote control, and gateway failures. Each log is bound to a precise timestamp and uses a cyclic overwrite storage mechanism, supporting retrieval and traceability via local ports and operation and maintenance mini-programs.

[0046] The cloud-based intelligent analysis module is the core computing unit for intelligent judgment of system equipment operating conditions, fault identification and classification, and prediction of hidden dangers. It receives standardized time-series datasets output by the edge computing processing module and, relying on built-in quantitative calculation models, fault judgment rules, risk scoring models, and time-series analysis algorithms, completes equipment output deviation calculation, equipment anomaly identification, fault classification and location, and long-term degradation trend warning. On the one hand, it outputs fault and warning data packets to the integrated application service module; on the other hand, it collects station-wide operation, fault, and maintenance work order data to form a training database, continuously iterating and optimizing judgment parameters and analysis models to reduce false alarms and missed alarms, adapting to the needs of unmanned, all-weather intelligent diagnosis in new energy power plants. The module integrates a data receiving and buffering subunit, an operating condition quantitative calculation subunit, a fault judgment comparison subunit, a fault classification and tracing subunit, a time-series trend prediction subunit, and a data output iteration subunit. Each subunit works in sequence to complete the full-link intelligent analysis and calculation.

[0047] The data receiving and buffering subunit is equipped with a data receiving interface for connecting to the edge computing processing module. Each standardized dataset is bound to a unique device number, a high-precision acquisition timestamp, a link status marker, and a local initial screening condition marker. The data receiving and buffering subunit is divided into five independent time-series buffer partitions according to device type: photovoltaic strings, inverters, transformer substations, environmental sensors, and video image metadata. The received standardized datasets are written into the corresponding time-series buffer queue according to device type, and the data in the queue is sorted in order according to the timestamp to generate a full-site time-series operating condition dataset.

[0048] The operating condition quantification calculation subunit is the core of equipment operating status quantification assessment. This subunit pre-constructs a theoretical standard output power database categorized by scenario, equipment model, and installed capacity. The database is stored in a multi-level classification based on solar irradiance range, site environmental temperature and humidity range, intraday operating period, and season, covering the standard output benchmark parameters of all mainstream equipment in centralized and distributed photovoltaic power plants. During actual operation, the operating condition quantification calculation subunit first retrieves the real-time output power of a single device, the corresponding site irradiance, and environmental temperature and humidity time-series data from the time-series cache queue of the data receiving cache subunit. Then, it matches this data with the theoretical standard output power of the same model of equipment under the same irradiance, temperature and humidity, and time period from the theoretical standard output power database. Finally, it calls the built-in power deviation rate quantification formula to complete the numerical calculation. The power deviation rate quantification formula is:

[0049] In the formula, The real-time power deviation rate of the equipment. This refers to the actual output power of the equipment. This provides the theoretical standard output power of the equipment under the same environment. After calculating the real-time power deviation rate, the real-time power deviation rate, theoretical standard output power, actual output power of the equipment, corresponding environmental sensor parameters, and data acquisition time sequence timestamps are bound and archived into a dedicated operating condition file for each individual device. This completely preserves the quantitative basis for the equipment's output deviation from the benchmark, providing quantitative indicator support for downstream fault diagnosis, power generation loss calculation, and string attenuation identification.

[0050] The fault determination and comparison subunit has a built-in set of parallel fault determination rules, which includes temperature threshold comparison rules, power deviation comparison rules, and communication link status identification rules. The temperature threshold comparison rules match the preset upper and lower temperature limits of the inverter, transformer substation, and photovoltaic string terminals, and compare them with the real-time temperature measurement values ​​of the equipment. The power deviation comparison rules retrieve the preset allowable power deviation rate threshold and compare it with the real-time power deviation rate output by the operating condition quantification calculation subunit. The communication link status identification rules read the link anomaly tag and heartbeat missing flag encapsulated in the data frame header. When any comparison result exceeds the corresponding threshold range, four types of differentiated anomaly tags are added to the corresponding equipment point: string anomaly, equipment overheating anomaly, power attenuation anomaly, and communication failure.

[0051] The fault grading and tracing subunit incorporates a fault urgency quantitative scoring model. This model uses five scoring dimensions: fault type, power deviation magnitude, equipment overheating duration, communication interruption duration, and estimated power generation loss. Each dimension is assigned an independent weight coefficient. Based on the urgency score range output by the model, faults are classified into three risk levels: high, medium, and low. The fault urgency score calculation logic is as follows:

[0052] In the formula, S represents the urgency score of the fault. , , , , and , respectively, are the weighting coefficients corresponding to power deviation, overheating duration, and estimated power generation loss. Power deviation rate, For the duration of continuous overheating of the equipment, The system estimates power generation losses during the fault period. A higher S score indicates greater power generation loss and equipment damage risk caused by the fault, and a higher priority for operation and maintenance. The fault classification and tracing subunit simultaneously traces the fault location, reads the unique number of the abnormal equipment and associates it with the site geographic coordinate library and the equipment branch topology library. It automatically binds the faulty equipment to the site, area, branch, precise geographic coordinates, and equipment level information. It encapsulates the fault type, risk level, geographic coordinates, out-of-limit index value, high-precision timestamp of fault triggering, and estimated power generation loss into a standardized fault analysis data package, which is then distributed synchronously to the downstream time-series trend prediction subunit and integrated application service module.

[0053] The time-series trend prediction subunit retrieves historical time-series operating condition files of the equipment stored in the cloud and constructs time-series trend curves of equipment output, temperature, voltage, and current on an hourly, daily, and monthly basis. The time-series trend prediction subunit has a built-in trend offset threshold and continuously monitors the slope of the trend curve. If the curve offset exceeds the preset trend offset threshold within multiple consecutive operating cycles, a hidden fault early warning and judgment data package carrying the fault development trend and the estimated fault outbreak time is generated.

[0054] The data output iteration subunit pushes standardized fault assessment data packages and latent fault early warning assessment data packages to the integrated application service module. The data output iteration subunit collects the time-series operation data of all station equipment, fault judgment records, maintenance archive work order data, and fault power generation loss data and stores them in the cloud model training database. It uses the collected business feedback data to update and correct the theoretical standard output power database, fault judgment threshold set, and fault urgency score weight coefficient, and completes the iterative update of the cloud analysis and judgment model.

[0055] The integrated application service module is the top-level business unit of the system facing the operation and maintenance end. It receives fault judgment and implicit early warning data packets output by the cloud intelligent analysis module, and integrates functions such as full-domain visualization display, hierarchical alarm push, intelligent work order dispatch, operation and maintenance closed-loop management, and model data feedback. It supports remote centralized operation and maintenance of new energy power plants with minimal human intervention. The module includes a full-domain visualization display sub-unit, a multi-channel hierarchical alarm push sub-unit, an intelligent generation and dispatch sub-unit for operation and maintenance work orders, a work order closed-loop management sub-unit, and a model data feedback sub-unit.

[0056] The full-domain visualization display subunit connects to the cloud-based intelligent analysis module, outputting a full-site time-series operating condition dataset and fault location tracing information, and incorporates multi-dimensional visualization rendering logic. The full-domain visualization display subunit is divided into a power station overview visualization zone, a single device parameter visualization zone, a regional map location visualization zone, a fault alarm statistics visualization zone, and a maintenance work order data statistics zone. The power station overview visualization zone displays global indicators such as installed capacity, cumulative power generation, daily / monthly / yearly power generation, power generation revenue, number of normal / abnormal devices, and number of communication faults. The single device parameter visualization zone displays real-time collected parameters of the three-phase grid voltage and daily power generation of each transformer substation, categorized by transformer substation. The regional map location visualization zone loads a geographic layer, using differentiated identifiers to mark centralized and distributed power stations, and employing green, red, yellow, and white markers to distinguish four types of equipment status points: normal operation, equipment failure, string abnormality, and communication failure. Clicking on a location retrieves the corresponding equipment's real-time operating parameters. Figure 3 As shown, the fault alarm statistics visualization displays the total number of devices, the number of abnormal devices, and the total number of high / medium / low risk alarms by region. It also shows the number of normal devices and the number of alarms at each risk level in each region. Figure 4As shown, the operation and maintenance work order data statistics are displayed in a page-by-page format, showing the work order number, work order type, work order status, faulty equipment, fault period, power loss due to fault, and number of times the work order has been handled.

[0057] The multi-channel hierarchical alarm push subunit is pre-configured with the mapping relationship between push channels and push objects corresponding to high, medium and low risk alarms. Push channels include cloud platform pop-ups, operation and maintenance mobile APP, and SMS. The multi-channel hierarchical alarm push subunit reads the fault risk level field in the standardized fault analysis data packet, matches the corresponding push channel, and encapsulates the fault location, fault type, out-of-limit indicators, and fault duration into an alarm message, which is then pushed synchronously to the corresponding operation and maintenance personnel's terminal.

[0058] The intelligent maintenance work order generation and dispatch subunit has built-in fault priority judgment logic, which retrieves the fault urgency score, the fault location's area, and the fault equipment type information from the standardized fault assessment data package. Based on the fault urgency score, the subunit sorts all abnormal points from high to low and automatically generates a dedicated maintenance work order for each abnormal point. The maintenance work order is automatically filled with the work order number, work order type, fault equipment, fault location, fault occurrence time, fault type, fault risk level, and estimated power generation loss, and simultaneously matches the standard maintenance tool list and standardized fault handling plan for the corresponding fault type. The subunit then distributes the work order to the corresponding maintenance handling unit according to the area to which the fault equipment belongs.

[0059] The work order closed-loop management subunit stores all dispatched maintenance work orders and provides entry points for updating work order statuses in two ways: "processing" and "completed". After maintenance personnel complete on-site fault handling, they can upload on-site handling photos, fill in the handling process, actual power loss, fault elimination status, and submit the work order for acceptance on the work order page. The work order closed-loop management subunit verifies the completeness of the work order acceptance submission information. After the verification is passed, the work order is marked as archived and completed, and all field information of the work order is fully retained to form a traceable maintenance archive.

[0060] The model data feedback subunit reads all the archived maintenance work orders and extracts the original fault assessment information, the actual causes of the fault on site, the fault handling time, the actual power generation loss data, and the equipment condition change data before and after the fault repair. The above maintenance business feedback data is then pushed synchronously to the cloud model training database of the cloud intelligent analysis module for iterative updates of the equipment condition analysis model, fault classification judgment model, and fault urgency score weight coefficient by the cloud intelligent analysis module.

[0061] The above are merely preferred embodiments of the invention and are not intended to limit the invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the invention should be included within the protection scope of the invention.

Claims

1. A smart power plant inspection, analysis, and monitoring system adapted for minimally staffed operation, characterized in that: include: The field sensing and acquisition module is used to deploy various types of field sensing and acquisition terminals to continuously collect raw operational data from all dimensions of the power plant's power generation equipment, power distribution equipment, and the station environment, achieving blind-spot-free sensing and acquisition of all station equipment. The network transmission module is used to build an integrated communication transmission link within the station, carrying bidirectional transparent transmission and unified access of front-end sensing and collection data; it monitors the operation status of the communication link in real time, identifies and marks link anomalies, and ensures stable and complete data transmission across levels. The edge computing processing module is deployed on the power plant edge gateway device. It is used to perform protocol parsing, data cleaning, noise filtering, and format standardization on the heterogeneous data collected from the front end. It also removes invalid and interference data through a steady-state judgment algorithm. It is configured with local data caching and breakpoint resume mechanism to complete the preliminary judgment of simple on-site working conditions. The cloud-based intelligent analysis module receives standardized datasets output by the edge computing processing module, performs quantitative calculations based on a preset equipment operating condition analysis model, automatically identifies various explicit and implicit abnormal operating conditions of the equipment, and completes fault classification, categorization, and source tracing. The integrated application service module is used to realize the visualization of power plant operation data, receive cloud-based fault analysis results and execute multi-channel hierarchical alarm push; generate operation and maintenance priorities based on fault urgency weights, automatically generate targeted operation and maintenance work orders and distribute them in a targeted manner; It supports closed-loop management of the entire process of operation and maintenance work order handling, acceptance, and archiving, and iterates and optimizes the cloud-based analysis and judgment model by feeding back operation and maintenance business data.

2. The new energy smart power station inspection, analysis and monitoring system adapted for minimal human intervention as described in claim 1, characterized in that, The sensing and acquisition terminal includes a photovoltaic string acquisition terminal, an inverter acquisition module, a transformer substation monitoring and control device, an environmental monitoring sensor, and on-site video acquisition equipment; The photovoltaic string acquisition terminal and the inverter acquisition module are used to acquire photovoltaic string voltage, current, output power, equipment temperature, load conditions, and communication heartbeat parameters. The transformer substation monitoring and control device is used to collect three-phase voltage, equipment temperature, load, and communication status parameters on the distribution side. The environmental monitoring sensor is used to collect environmental parameters such as irradiance, temperature, and humidity at the site. The on-site video acquisition equipment is used to collect real-time video data of the equipment at the site.

3. The new energy smart power station inspection, analysis and monitoring system adapted for minimal human intervention as described in claim 1, characterized in that, The edge gateway device has a built-in link monitoring and processing unit, which includes a heartbeat message sending and receiving subunit, a data packet statistics and verification subunit, a link status storage subunit, and an anomaly identification encapsulation subunit.

4. The new energy smart power station inspection, analysis and monitoring system adapted for minimal human intervention as described in claim 3, characterized in that, The heartbeat message transceiver subunit is pre-configured with the heartbeat transmission period for each communication link. For each wired Ethernet, fiber optic, 4G / 5G, or LoRa wireless link, it periodically generates a heartbeat detection message carrying a unique link identifier and sends it to the collection terminal at the end of the link. After receiving the heartbeat detection message, the terminal immediately sends back a response message carrying the corresponding link identifier and local receiving timestamp; The heartbeat message transceiver subunit records the sending time of each probe message and the receiving time of the response message, calculates the round-trip transmission delay of a single message, and continuously counts the number of heartbeat probes that have not received a response message within a preset period. The data packet statistics and verification subunit adds a continuously increasing frame sequence number to the service data frames transmitted through each link, performs CRC cyclic redundancy check on each service data frame, and counts the total number of data frames sent by the gateway and the number of valid verification data frames returned by the terminal in each link per unit time. The real-time packet loss rate of the link is calculated by combining the total number of frames and the number of valid received frames. Simultaneously collect wireless link RSSI signal strength and wired link port bandwidth occupancy indicators; The link state storage subunit establishes a link state data table in local non-volatile memory. The data table fields include link number, link type, real-time round-trip delay, periodic packet loss rate, number of missing heartbeat responses, wireless signal strength, and link normal / abnormal status flag bit. The link state storage subunit writes the real-time indicators output by the heartbeat message sending and receiving subunit and the data packet statistics and verification subunit into the link state data table in real time; The anomaly identification encapsulation subunit has a built-in set of link anomaly judgment thresholds, which includes the link delay threshold, packet loss rate threshold, consecutive heartbeat no response count threshold, and wireless signal strength lower limit threshold. The anomaly identification encapsulation subunit reads each link indicator in the link status data table in real time and compares the real-time indicators with the corresponding thresholds one by one. When any indicator exceeds the threshold range, a link anomaly label is generated. The anomaly label, the anomaly indicator value, and the anomaly trigger timestamp are encapsulated into the additional field of the header of all service data frames currently transmitted on the link to complete the link anomaly status marking. For links that have not triggered threshold exceedance, a link normal identification field is appended to the header of the service data frame.

5. The new energy smart power station inspection, analysis and monitoring system adapted for minimal human intervention as described in claim 1, characterized in that, The edge gateway device also has a built-in data preprocessing unit, which includes a data cleaning subunit, a data deduplication subunit, a local storage subunit, and a simplified operating condition discrimination subunit.

6. The new energy smart power station inspection, analysis and monitoring system adapted for minimal human intervention as described in claim 5, characterized in that, The data cleaning subunit receives real-time power sampling values ​​from the device and stores each real-time power sampling value into a sliding window memory queue. The queue follows a first-in, first-out (FIFO) rule, and when the window is full of new data, the oldest historical sampling data stored in the queue is automatically removed. The data cleaning subunit reads all valid power sampling values ​​within the sliding window in real time and calculates the arithmetic mean of all sampling values ​​within the window to obtain the short-time average power. The data cleaning subunit reads locally preset power fluctuation thresholds, which are stored according to device type and installed capacity, distinguishing independent threshold parameters for photovoltaic strings, inverters, and transformer substations. The data cleaning subunit calculates the absolute difference between the real-time power and the short-time average power and compares the calculated difference with the corresponding device's allowable fluctuation threshold. When the absolute difference between the real-time power and the short-time average power is greater than the preset allowable fluctuation threshold, the current sampled data is determined to be jump interference data and the sampled data is directly discarded; when the absolute difference between the real-time power and the short-time average power is less than or equal to the preset allowable fluctuation threshold, the current sampled data is determined to be valid steady-state data, the sampled data is retained and transmitted to the downstream data deduplication subunit. The data deduplication subunit receives the valid data output by the steady-state determination, uses the device's unique number and high-precision timestamp as joint verification identifiers, compares multiple data reports repeatedly submitted by the same device at the same timestamp, retains only the first original sampled data, and deletes the remaining duplicate messages. The local storage subunit is divided into two independent storage blocks according to data type: real-time data cache partition and offline backup cache partition. The real-time data cache partition is used to temporarily store real-time sampled data under normal transmission conditions, while the offline backup cache partition is specifically used to store all original standardized data collected on-site during offline periods when the communication link is disconnected. The local storage unit binds a time-series timestamp, a unique identifier of the corresponding acquisition device, and a data generation sequence number to each piece of cached data, and synchronously records the data cache write time. The local storage subunit has built-in breakpoint resume scheduling logic, continuously reading the link connectivity status flag field pushed by the network transmission module in real time; when the link status is marked as disconnected, it stops pushing data to the upstream cloud and continuously writes newly generated standardized data into the offline backup cache partition. When the link status flag is restored to normal connectivity, the breakpoint resume scheduling logic reads all cached data in the offline backup cache partition, sorts the data from smallest to largest according to the time sequence timestamp bound to each data, and resends the cached data to the cloud intelligent analysis layer in batches and segments according to the time sequence. After all offline cached data is resent, it switches to real-time synchronization push mode. The simplified operating condition discrimination subunit pre-stores a set of basic threshold parameters for different equipment types. The threshold parameter sets are configured with voltage thresholds, current thresholds, power thresholds, and temperature thresholds for photovoltaic strings, inverters, and transformer substations, respectively. The simplified operating condition discrimination subunit reads the standardized data after data preprocessing and format normalization one by one, extracts the equipment measurement point values ​​in the data, and compares them one by one with the basic threshold parameters that match the corresponding equipment. When the measurement point value exceeds the upper or lower limit range of the corresponding threshold, an on-site anomaly preliminary screening mark field is added to this data. When the measurement point value is within the threshold range, a normal operation mark field is added to this data.

7. The new energy smart power station inspection, analysis and monitoring system adapted for minimal human intervention as described in claim 1, characterized in that, The cloud-based intelligent analysis module includes a data receiving and caching subunit, a working condition quantification and calculation subunit, a fault judgment and comparison subunit, a fault classification and tracing subunit, a time-series trend prediction subunit, and a data output iteration subunit.

8. The new energy smart power station inspection, analysis and monitoring system adapted for minimal human intervention as described in claim 7, characterized in that, The data receiving and buffering subunit is equipped with a data receiving interface for the edge computing processing module. Each standardized dataset is bound to a unique device number, a high-precision acquisition timestamp, a link status marker, and a local initial screening condition marker. The data receiving and buffering subunit is divided into five independent time-series buffer partitions according to device type: photovoltaic strings, inverters, transformer substations, environmental sensors, and video image metadata. The received standardized datasets are written into the corresponding time-series buffer queues according to device type, and the data in the queues is sorted in order according to the timestamp to generate a full-site time-series operating condition dataset. The operating condition quantization calculation subunit pre-stores a theoretical standard output power database matching different installed capacities and equipment models. The theoretical standard output power database stores theoretical power parameters according to irradiation range, ambient temperature and humidity range, and operating period. The operating condition quantization calculation subunit retrieves the real-time output power data of the equipment in the time-series cache queue, calls the built-in power deviation rate quantization formula to complete the numerical calculation, outputs the real-time power deviation rate of a single equipment, and binds the deviation rate value, theoretical standard output power, actual output power of the equipment, and corresponding environmental sensor parameters into the equipment operating condition file. The fault determination and comparison subunit has a built-in set of parallel fault determination rules, which includes temperature threshold comparison rules, power deviation comparison rules, and communication link status identification rules. The temperature threshold comparison rules match the preset upper and lower temperature limits of the inverter, transformer substation, and photovoltaic string terminals, and compare them with the real-time temperature measurement values ​​of the equipment. The power deviation comparison rules retrieve the preset allowable power deviation rate threshold and compare it with the real-time power deviation rate output by the operating condition quantification calculation subunit. The communication link status identification rules read the link anomaly tag and heartbeat missing tag encapsulated in the data frame header. When any comparison result exceeds the corresponding threshold range, four types of differentiated anomaly tags are added to the corresponding equipment point: string anomaly, equipment overheating anomaly, power attenuation anomaly, and communication failure. The fault classification and tracing subunit has a built-in fault emergency quantitative scoring model. It uses fault type, power deviation range, equipment overheating duration, communication interruption duration, and estimated power generation loss as scoring dimensions and configures corresponding weight coefficients. Based on the scoring value range, it divides the fault risk level into three levels: high, medium, and low. The fault classification and tracing subunit binds the unique number of the abnormal equipment, the geographical coordinates of the site, the branch information to which the equipment belongs, and the fault trigger timestamp to complete the fault location tracing and positioning. It encapsulates the fault type, risk level, location coordinates, out-of-limit index value, and fault trigger time to generate a standardized fault analysis data package. The time-series trend prediction subunit retrieves historical time-series operating condition files of the equipment stored in the cloud and constructs time-series change trend curves of equipment output, temperature, voltage, and current with hourly, daily, and monthly periods, respectively. The time-series trend prediction subunit has a built-in trend offset threshold and continuously monitors the change in the slope of the trend curve. If the curve offset exceeds the preset trend offset threshold in multiple consecutive operating cycles, a hidden fault early warning and judgment data package carrying the fault development trend and the estimated fault outbreak time is generated. The data output iteration subunit pushes standardized fault assessment data packages and latent fault early warning assessment data packages to the integrated application service module; the data output iteration subunit collects the time-series operation data of all station equipment, fault judgment records, maintenance archive work order data, and fault power generation loss data and stores them in the cloud model training database. It uses the collected business feedback data to update and correct the theoretical standard output power database, fault judgment threshold set, and fault urgency score weight coefficient, and completes the iterative update of the cloud analysis and judgment model.

9. The new energy smart power station inspection, analysis and monitoring system adapted for minimal human intervention as described in claim 1, characterized in that, The integrated application service module includes a full-domain visualization display subunit, a multi-channel hierarchical alarm push subunit, an intelligent generation and dispatch subunit for operation and maintenance work orders, a work order closed-loop management subunit, and a model data feedback subunit.

10. The new energy smart power station inspection, analysis and monitoring system adapted for minimal human intervention as described in claim 9, characterized in that, The full-domain visualization display sub-unit connects to the cloud-based intelligent analysis module to output the full-site time-series operating condition dataset and fault location tracing information, and has built-in multi-dimensional visualization rendering logic; The full-domain visualization display sub-unit is divided into a power plant overview visualization zone, a single device parameter visualization zone, a regional map location visualization zone, a fault alarm statistics visualization zone, and a maintenance work order data statistics zone. The power plant overview visualization section displays global indicators such as installed capacity, cumulative power generation, daily / monthly / yearly power generation, power generation revenue, number of normal / abnormal devices, and number of communication failures. The single device parameter visualization section displays real-time collected parameters of three-phase grid voltage and daily power generation of each transformer substation. The regional map point visualization section loads a geographic layer and uses differentiated identifiers to mark centralized and distributed power plants. Green, red, yellow, and white markers are used to distinguish four types of equipment status points: normal operation, equipment failure, string abnormality, and communication failure. Clicking on a point allows you to retrieve the real-time operating parameters of the corresponding equipment. The fault alarm statistics visualization section counts the total number of devices, the number of abnormal devices, and the total number of high / medium / low risk alarms. It displays the number of normal devices and the number of alarms at each level in each region. The operation and maintenance work order data statistics are displayed in a divided and paginated manner, showing the work order number, work order type, work order status, faulty equipment, fault period, power loss due to fault, and number of times the work order has been handled, providing a complete list of work order data. The multi-channel hierarchical alarm push subunit is pre-configured with a mapping relationship between push channels and push objects corresponding to high, medium and low risk alarms. The push channels include cloud platform pop-ups, operation and maintenance mobile APP, and SMS. The multi-channel hierarchical alarm push subunit reads the fault risk level field in the standardized fault analysis data packet, matches the corresponding push channel, and encapsulates the fault location, fault type, out-of-limit indicators, and fault duration into an alarm message, which is then synchronously pushed to the corresponding operation and maintenance personnel's terminal. The intelligent maintenance work order generation and dispatch subunit incorporates fault priority determination logic, retrieving fault urgency scores, fault location regions, and fault equipment type information from standardized fault assessment data packets. Based on the fault urgency scores, the subunit sorts all abnormal locations from highest to lowest, automatically generating a dedicated maintenance work order for each abnormal location. The maintenance work order is automatically populated with the work order number, work order type, fault equipment, fault location, fault occurrence time, fault type, fault risk level, and estimated power generation loss, simultaneously matching the standard maintenance tool list and standardized fault handling plan for the corresponding fault type. Finally, the subunit distributes the work order to the corresponding regional maintenance handling unit according to the region to which the fault equipment belongs. The work order closed-loop management subunit stores all dispatched maintenance work orders and provides entry points for updating work order statuses in progress and completed. After completing on-site fault handling, maintenance personnel can upload on-site handling photos, fill in the handling process, actual power loss, fault elimination status, and submit the work order for acceptance on the work order page; the work order closed-loop management sub-unit verifies the completeness of the work order acceptance submission information, and marks the work order as archived and completed after the verification is passed, and retains all field information of the work order to form a traceable maintenance file; The model data feedback subunit reads all the archived maintenance work orders, extracts the original fault assessment information, the actual causes of the fault on site, the fault handling time, the actual power generation loss data, and the equipment operating condition change data before and after the fault repair, and pushes the above maintenance business feedback data to the cloud model training database of the cloud intelligent analysis module. This data is used by the cloud intelligent analysis module to iteratively update the equipment operating condition analysis model, the fault classification judgment model, and the fault urgency score weight coefficient.

Images