New energy power station perception control method and system

Abstract

The invention relates to the technical field of new energy power station control, and particularly discloses a new energy power station perception control method and system, and the method comprises the steps: S1, data collection, S2, state perception, S3, monitoring and early warning, S4, decision and optimization control, S5, execution and feedback, S6, intelligent operation and maintenance, and S7, self-healing scheduling. The deep neural network DNN is utilized to process a large amount of sensor data, so that the system can learn modes from complex historical data, accurately identify conditions such as power fluctuation and voltage abnormity, and timely discover potential abnormity through real-time updating of the model, so that the possibility of misinformation and missing report is reduced, and the reliability of the system is improved. The system monitors the operation state of the power station in real time, can quickly recognize and classify abnormal conditions, provides instant data support for decision making, can take measures in advance through abnormal prediction, prevents problems from being further expanded, and can recognize equipment needing to be maintained and prevents faults from being expanded through patrol inspection and prediction analysis of the unmanned aerial vehicle.

Description

Technical Field

[0001] The present invention belongs to the technical field of new energy power station control, and specifically relates to a new energy power station sensing control method and system. Background Art

[0002] In the operation and management of new energy power stations, traditional perception control methods mainly rely on preset threshold judgments and regular manual inspections. These methods are powerless in the face of complex and changeable power station operating environments and massive sensor data. The preset thresholds are often difficult to accurately reflect subtle changes in the power station's operating status, resulting in frequent false alarms and missed alarms, which not only increases operation and maintenance costs, but also makes it difficult to discover and deal with key problems in a timely manner.

[0003] Traditional inspection methods rely on manual operations, which are not only inefficient, but also difficult to cover all key areas and details, especially in bad weather or remote areas, which significantly increases the difficulty and cost of inspections. In addition, manual inspections are often subjective and delayed, making it difficult to detect potential problems in a timely and accurate manner.

[0004] In this regard, the inventors propose a new energy power station perception control method and system to solve the above problems. Summary of the invention

[0005] The purpose of the present invention is to provide a new energy power station perception control method and system to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A new energy power station sensing control method comprises the following steps:

[0008] S1. Data collection: arrange multi-source sensor networks, collect parameters of new energy equipment, deploy edge computing nodes, preliminarily process and filter the data collected by sensors, reduce redundant data transmission, obtain external data such as weather, load forecast, and market electricity prices in real time, and integrate them with historical data to form comprehensive data input;

[0009] S2. Status perception: real-time monitoring of the operating status of power generation units, energy storage systems, and transmission and distribution network subsystems to ensure comprehensive data coverage. The adaptive health index method is used to evaluate the health status of equipment, predict the probability of failure, and estimate the remaining service life of equipment through big data analysis and machine learning algorithms.

[0010] S3, monitoring and early warning, using deep learning algorithms to monitor the operation status of the power station in real time, identify power fluctuations and voltage anomalies, design multi-level response strategies, and formulate different response plans according to the severity of the anomaly, including automatic adjustment for low-level anomalies, automatic adjustment for medium-level anomalies and notification of operation and maintenance, and manual intervention for high-level anomalies;

[0011] S4, decision-making and optimization control, using intelligent energy scheduling, combined with real-time load demand, energy storage status, and weather forecast factors to optimize energy distribution and use, and using reinforcement learning algorithms to optimize power station operating parameters in real time, including power factor and conversion efficiency;

[0012] S5, Execution and feedback. Each subsystem in the power station has autonomous control capabilities. When the central system fails, each subsystem can work together to maintain the basic operating status of the system, establish a closed-loop feedback mechanism, continuously monitor the execution effect of control instructions, and feed back the results to the decision-making layer. The control strategy is automatically adjusted according to the feedback data to improve the system's response speed and decision-making accuracy;

[0013] S6. Intelligent operation and maintenance: Develop an intelligent operation and maintenance platform that can automatically generate maintenance plans, and use remote operation and maintenance robots to detect and repair equipment. Use drone inspections to regularly inspect the surface cleanliness of photovoltaic panels and the wear of wind turbine blades to ensure that the equipment is in the best condition.

[0014] S7, self-healing scheduling, when a local fault is detected, the fault area is automatically isolated to prevent the impact from spreading to the entire power station. When a potential problem is detected, the system automatically adjusts the operating strategy until the fault is completely repaired.

[0015] Preferably, the multi-source sensor network includes photovoltaic power station sensors, wind power generation sensors, energy storage system sensors, environmental monitoring sensors, grid access monitoring sensors and safety protection sensors.

[0016] Preferably, the comprehensive data in step S1 is obtained by data fusion, and its expression is:

[0017] Dt=α·Dr+β·Dh+γ·Df

[0018] Where Dt: the fused data set, used for subsequent analysis and decision-making;

[0019] Dr: Data collected in real time;

[0020] Dh: historical data, reflecting the past behavior and trends of the system;

[0021] Df: forecast data, including weather forecast and load forecast;

[0022] α, β, γ: weight coefficients, reflecting the contribution of various types of data to the final fusion result, and satisfying α+β+γ=1;

[0023] By integrating data from different sources, the reliability and accuracy of input data are improved and the errors caused by a single data source are reduced.

[0024] Preferably, in step S2, the adaptive health index AHi calculates the current health status of the device by standardizing various monitoring data, and can detect potential problems in a timely manner. The expression is:

[0025]

[0026] Where AH i: health index of equipment or system, the higher the value, the worse the health status;

[0027] Xj: jth monitoring data, including temperature and voltage;

[0028] μj: the historical average value of the j-th data;

[0029] σj: standard deviation of the j-th data;

[0030] n: the number of monitoring data.

[0031] Preferably, in step S3, a multi-level response strategy is adopted in the deep learning algorithm to automatically select the most appropriate response measure according to the severity of the abnormality, and its expression is:

[0032]

[0033] in The response strategy that the system should adopt for the currently detected anomaly;

[0034] P(c1|X): probability of low-level anomaly;

[0035] P(c2|X): probability of medium-level anomaly;

[0036] P(c3|X): probability of high-level anomaly;

[0037] θ1, θ2: probability thresholds that determine the switching point of the response strategy.

[0038] Preferably, the intelligent energy scheduling in step S4 is expressed as:

[0039]

[0040] Po: optimal scheduling power allocation;

[0041] Pi: power output of the i-th energy unit;

[0042] C(Pi): the power cost function of the i-th energy unit, which is usually related to the power generation cost and energy consumption;

[0043] R(Pi): Power volatility or risk index, measuring the impact of uncertainty on the system;

[0044] λ: Risk preference coefficient, reflecting the system's tolerance to volatility risk;

[0045] m: the number of energy units;

[0046] By optimizing power distribution, operating costs are reduced while fluctuation risks are controlled and stability is improved.

[0047] Preferably, in step S4, the reinforcement learning algorithm gradually approaches the optimal strategy by continuously updating the value function Q(s,a) to achieve dynamic optimization control, specifically:

[0048]

[0049] Where Q(s,a): the value function of taking action a in state s;

[0050] α: learning rate, which determines the update step size;

[0051] r: immediate reward, reflecting the direct benefit of the current action;

[0052] γ: discount factor, measuring the current value of future earnings;

[0053] s′: the new state the system transfers to after executing action a;

[0054] a′: action in the new state s′;

[0055] In the dynamic environment of a new energy power station, the above formula can help the system continuously optimize the control strategy and improve the overall operating efficiency in a complex and changing environment.

[0056] Preferably, the expression for isolating the fault in step S7 is:

[0057]

[0058] Where I(t): fault isolation index, used to evaluate whether a fault area needs to be isolated;

[0059] Fk(t): the value of the kth fault indicator at time t;

[0060] wk: weight of the kth fault indicator, reflecting its importance;

[0061] K: the number of fault indicators; the fault isolation algorithm determines whether the fault area needs to be isolated by integrating the values ​​of multiple fault indicators to ensure the stable operation of the system.

[0062] A new energy power station sensing control system, comprising:

[0063] The data processing and analysis module includes a data acquisition and preprocessing unit and a data analysis and anomaly detection unit. It performs denoising and standardization on the collected sensor data, analyzes and predicts the operating status of the new energy power station by training a deep neural network model, identifies abnormal situations, including power fluctuations and voltage anomalies, and calculates the health index of the equipment to evaluate the current status and expected life of the equipment.

[0064] The control and decision-making module includes an energy management unit and a multi-level response strategy unit. The energy management unit dynamically adjusts the energy production and distribution strategy of the power station according to real-time data, optimizes the operation of photovoltaic panels and energy storage systems, and adjusts the wind turbine angle, photovoltaic panel angle and working state of the energy storage system in real time to maximize energy output and extend equipment life. The multi-level response strategy unit includes low-level response, medium-level response and high-level response;

[0065] Intelligent operation and maintenance module, which predicts potential equipment failures based on historical data and deep learning models, plans maintenance and repair work in advance to avoid sudden failures, and uses drones to regularly inspect photovoltaic panels and other key equipment, capture image and video data, and identify damage or contamination;

[0066] The feedback and self-healing module includes a closed-loop feedback control unit and a self-healing function unit. It provides real-time feedback on the execution effect of the control strategy and dynamically adjusts the control plan based on the feedback data to ensure that the power station is always in the best operating state. When equipment failure or operation deviation is detected, the system automatically adjusts the operating parameters, implements the self-healing function, restores the equipment to normal state, and reduces downtime and operation risks.

[0067] Preferably, the low-level response: automatically adjusts operating parameters to cope with minor anomalies to ensure continuous and stable operation of the power station;

[0068] Medium-level response: When a moderate anomaly is detected, the system automatically adjusts the equipment and notifies the operation and maintenance personnel to conduct an inspection.

[0069] High-level response: For serious anomalies, the system performs an emergency shutdown and issues an alarm, requiring human intervention to prevent the accident from spreading.

[0070] Compared with the prior art, the present invention has the following beneficial effects:

[0071] (1) The present invention uses a deep neural network (DNN) to process a large amount of sensor data, enabling the system to learn patterns from complex historical data and accurately identify power fluctuations, voltage anomalies, etc. Through real-time updates of the model, the system can promptly detect potential anomalies and reduce the possibility of false alarms and missed alarms. The system monitors the operating status of the power station in real time, can quickly identify and classify abnormal situations, and provide real-time data support for decision-making. By predicting anomalies, the system can take measures in advance to prevent the problem from further expanding.

[0072] (2) Through drone inspection and predictive analysis, the present invention can identify equipment that needs maintenance and prevent faults from expanding. When a fault is detected, the system can automatically isolate the faulty part to protect other equipment from being affected, thereby ensuring the overall safety of the power plant. When a problem is detected, the system automatically adjusts operating parameters to reduce the impact on power plant operations, and achieves self-healing through continuous monitoring to prevent small problems from developing into major faults.

[0073] (3) The present invention adopts a hierarchical response scheme based on the severity of the abnormality, from automatic adjustment to manual intervention, step by step. Low-level abnormalities can be automatically repaired by the system to avoid delays in manual intervention; high-level abnormalities are promptly notified to the operation and maintenance personnel for processing to prevent serious failures from occurring. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 This is a flow chart of a new energy power station sensing control method of the present invention;

[0075] Figure 2 It is a block diagram of the multi-source sensor network of the present invention;

[0076] Figure 3 This is a block diagram of a new energy power station sensing control system of the present invention;

[0077] Figure 4 It is a block diagram of the data processing and analysis module of the present invention;

[0078] Figure 5 It is a block diagram of the control and decision-making module of the present invention;

[0079] Figure 6 This is a block diagram of the feedback and self-healing module of the present invention. DETAILED DESCRIPTION

[0080] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0081] Embodiment 1:

[0082] See also Figure 1 and Figure 2 As shown, a new energy power station sensing control method includes the following steps:

[0083] S1. Data collection: arrange multi-source sensor networks, collect parameters of new energy equipment, deploy edge computing nodes, preliminarily process and screen the data collected by sensors, reduce redundant data transmission, and improve data processing efficiency. Obtain weather, load forecast, and market electricity price external data in real time, and integrate them with historical data to form comprehensive data input;

[0084] S2. Status perception: real-time monitoring of the operating status of power generation units, energy storage systems, and transmission and distribution network subsystems to ensure comprehensive data coverage. The adaptive health index method is used to evaluate the health status of equipment, predict the probability of failure, and estimate the remaining service life of equipment through big data analysis and machine learning algorithms.

[0085] S3, monitoring and early warning, using deep learning algorithms to monitor the operation status of the power station in real time, identify power fluctuations and voltage anomalies, design multi-level response strategies, and formulate different response plans according to the severity of the anomaly, including automatic adjustment for low-level anomalies, automatic adjustment for medium-level anomalies and notification of operation and maintenance, and manual intervention for high-level anomalies;

[0086] S4, decision-making and optimization control, using intelligent energy scheduling, combined with real-time load demand, energy storage status, and weather forecast factors to optimize energy distribution and use, and using reinforcement learning algorithms to optimize power station operating parameters in real time, including power factor and conversion efficiency;

[0087] S5, Execution and feedback. Each subsystem in the power station has autonomous control capabilities. When the central system fails, each subsystem can work together to maintain the basic operating status of the system, establish a closed-loop feedback mechanism, continuously monitor the execution effect of control instructions, and feed back the results to the decision-making layer. The control strategy is automatically adjusted according to the feedback data to improve the system's response speed and decision-making accuracy;

[0088] S6. Intelligent operation and maintenance: Develop an intelligent operation and maintenance platform that can automatically generate maintenance plans, and use remote operation and maintenance robots to detect and repair equipment. Use drone inspections to regularly inspect the surface cleanliness of photovoltaic panels and the wear of wind turbine blades to ensure that the equipment is in the best condition.

[0089] S7, self-healing scheduling, when a local fault is detected, the fault area is automatically isolated to prevent the impact from spreading to the entire power station. When a potential problem is detected, the system automatically adjusts the operating strategy until the fault is completely repaired.

[0090] Specifically, the multi-source sensor network includes photovoltaic power station sensors, wind power generation sensors, energy storage system sensors, environmental monitoring sensors, grid access monitoring sensors and safety protection sensors;

[0091] The photovoltaic power station sensors include light sensors / radiometers, temperature sensors, and voltage and current sensors. Light sensors / radiometers measure the irradiance of sunlight, that is, the light power density per unit area, and obtain light intensity data in real time to help evaluate the power generation efficiency of photovoltaic cells, optimize the angle adjustment of photovoltaic arrays, and ensure the best power generation conditions.

[0092] Temperature sensors monitor the temperature of photovoltaic modules and the surrounding environment. Temperature directly affects the conversion efficiency of photovoltaic modules. The data from temperature sensors are used to adjust the operating conditions of photovoltaic cells, such as starting and stopping the cooling system, to improve power generation efficiency and prevent overheating damage.

[0093] Voltage and current sensors monitor the voltage and current output of the PV array. These sensors help understand the power generation status in real time and detect abnormal conditions such as short circuits, open circuits or other electrical faults to ensure the stable operation of the power station.

[0094] Wind power generation sensors, including anemometers, wind direction sensors and vibration sensors. Anemometers measure wind speed and provide basic data for wind power generation. Wind speed data is used to adjust the operating status of wind turbines, such as the angle and speed of blades, to maximize power generation while avoiding damage to equipment under extreme wind speeds.

[0095] Wind direction sensor, which measures the direction of the wind. The wind direction data is used to adjust the direction of the wind turbine so that it always faces the wind source to improve the utilization rate of wind energy;

[0096] Vibration sensors monitor the vibration of key components such as wind turbine towers and blades. Vibration sensors can detect potential mechanical problems in advance, such as bearing wear or blade imbalance, prevent large-scale failures, and improve the safety and life of equipment.

[0097] Energy storage system sensors, including battery temperature sensors, battery voltage and current sensors, and status monitoring sensors. Battery temperature sensors monitor the temperature of energy storage battery packs. Temperature is critical to battery performance. Too high or too low temperatures will affect the battery's energy storage efficiency and life. Temperature sensors ensure that the battery operates within the optimal temperature range and prolongs battery life.

[0098] Battery voltage and current sensors measure the voltage and current of the energy storage battery pack, monitor the battery charge and discharge status in real time, prevent abnormal conditions such as overcharge and overdischarge, protect the battery and ensure the stability of the energy storage system;

[0099] Condition monitoring sensors monitor the health status of the energy storage system, such as battery capacity and internal resistance. By monitoring the health status of the battery, the battery service life and the time when maintenance is required are predicted, and the scheduling strategy of the battery management system BMS is optimized.

[0100] Environmental monitoring sensors, including weather station sensors and dust sensors. Weather station sensors comprehensively measure environmental factors such as temperature, humidity, rainfall, wind speed, wind direction, etc., provide detailed environmental data, help analyze the impact of the external environment on the power generation system, optimize operation strategies, and improve the power generation efficiency and safety of the overall power station.

[0101] Dust sensors monitor the dust concentration in the air, which can help determine the cleanliness of the photovoltaic panel surface and arrange cleaning in time to avoid affecting the light absorption efficiency of the photovoltaic panel;

[0102] Grid access monitoring sensors, including power quality sensors and harmonic sensors, monitor the power quality of the grid connection point, including voltage, current, frequency, power factor and other parameters, to ensure that the power output of the power station meets the grid standards and prevent grid connection failure or fines caused by power quality problems;

[0103] Harmonic sensors detect the harmonic content in the power grid, identify and control the harmonic pollution of the power station to the power grid, ensure that the operation of the power station meets the compatibility requirements of the power grid, and prevent equipment damage caused by harmonics.

[0104] Safety protection sensors, including smoke sensors and leakage sensors. Smoke sensors detect smoke in power station facilities to prevent fires. When signs of fire appear, they can promptly alarm and initiate corresponding safety protection measures to avoid major accidents.

[0105] Leakage sensors monitor the leakage of the system, detect and handle leakage problems in a timely manner, and ensure the safety of equipment and personnel.

[0106] The sensor network in the above-mentioned new energy power station monitors key parameters in real time in different subsystems and feeds data back to the central control system. These sensors not only ensure the safe and stable operation of the power station, but also improve the power generation efficiency and equipment service life of the power station through data collection and analysis. In addition, environmental monitoring sensors also provide the necessary external condition data for optimizing system operation.

[0107] Specifically, the comprehensive data in step S1 is obtained through data fusion, and its expression is:

[0108] Dt=α·Dr+β·Dh+γ·Df

[0109] Where Dt: the fused data set, used for subsequent analysis and decision-making;

[0110] Dr: Data collected in real time;

[0111] Dh: historical data, reflecting the past behavior and trends of the system;

[0112] Df: forecast data, including weather forecast and load forecast;

[0113] α, β, γ: weight coefficients, reflecting the contribution of various types of data to the final fusion result, and satisfying α+β+γ=1;

[0114] By integrating data from different sources, the reliability and accuracy of input data are improved and the errors that may be caused by a single data source are reduced. The above formula helps the system to comprehensively consider real-time, historical and predicted data when making decisions, making the decisions more comprehensive and reasonable.

[0115] Specifically, in step S2, the adaptive health index calculates the current health status of the device by standardizing various monitoring data, and can detect potential problems in a timely manner. The expression is:

[0116]

[0117] Where Hi: health index of the device or system, the higher the value, the worse the health status;

[0118] Xj: jth monitoring data, including temperature and voltage;

[0119] μj: the historical average value of the j-th data;

[0120] σj: standard deviation of the j-th data;

[0121] n: the number of monitoring data;

[0122] The health index can help operation and maintenance personnel prioritize devices with higher health indexes and prevent minor problems from developing into major failures.

[0123] Specifically, in step S3, the deep learning algorithm adopts a multi-level response strategy to automatically select the most appropriate response measure according to the severity of the abnormality, and its expression is:

[0124]

[0125] Where R(y^): the response strategy that the system should take for the currently detected anomaly;

[0126] P(c1|X): probability of low-level anomaly;

[0127] P(c2|X): probability of medium-level anomaly;

[0128] P(c3|X): probability of high-level anomaly;

[0129] θ1, θ2: probability thresholds, which determine the switching point of the response strategy;

[0130] This formula sets corresponding response strategies for different types of anomalies to ensure that the system can respond to abnormal situations quickly and reasonably, prevent the problem from expanding or the system from being damaged. Through multi-level response strategies, the most appropriate response measures are automatically selected to reduce unnecessary downtime and improve the reliability and economy of the system.

[0131] Specifically, the intelligent energy scheduling in step S4 is expressed as:

[0132]

[0133] Po: optimal scheduling power allocation;

[0134] P i: power output of the i-th energy unit;

[0135] C(Pi): the power cost function of the i-th energy unit, which is usually related to the power generation cost and energy consumption;

[0136] R(Pi): Power volatility or risk index, measuring the impact of uncertainty on the system;

[0137] λ: Risk preference coefficient, reflecting the system's tolerance to volatility risk;

[0138] m: the number of energy units;

[0139] By optimizing power distribution, reducing system operating costs, controlling fluctuation risks and improving system stability, intelligent energy scheduling can reasonably allocate the output of each unit when multiple new energy sources coexist, ensuring the economy and reliability of the system.

[0140] Specifically, in step S4, the reinforcement learning algorithm gradually approaches the optimal strategy by continuously updating the value function Q(s,a) to achieve dynamic optimization control, specifically:

[0141]

[0142] Where Q(s,a): the value function of taking action a in state s;

[0143] α: learning rate, which determines the update step size;

[0144] r: immediate reward, reflecting the direct benefit of the current action;

[0145] γ: discount factor, measuring the current value of future earnings;

[0146] s′: the new state the system transfers to after executing action a;

[0147] a′: action in the new state s′;

[0148] In the dynamic environment of a new energy power station, the above formula can help the system continuously optimize the control strategy and improve the overall operating efficiency in a complex and changing environment.

[0149] Specifically, the expression for isolating the fault in step S7 is:

[0150]

[0151] Where I(t): fault isolation index, used to evaluate whether a fault area needs to be isolated;

[0152] Fk(t): the value of the kth fault indicator at time t;

[0153] wk: weight of the kth fault indicator, reflecting its importance;

[0154] K: the number of fault indicators; the fault isolation algorithm determines whether it is necessary to isolate the fault area by integrating the values ​​of multiple fault indicators to ensure the stable operation of the system; the algorithm can respond quickly when a system fault occurs, minimize the impact of the fault on the system, and ensure the continuity and reliability of the system.

[0155] As can be seen above, using deep neural networks (DNN) to process a large amount of sensor data enables the system to learn patterns from complex historical data and accurately identify power fluctuations, voltage anomalies, etc. Through real-time updates of the model, the system can promptly detect potential anomalies and reduce the possibility of false alarms and missed alarms. The system monitors the operation status of the power station in real time, can quickly identify and classify abnormal situations, and provide real-time data support for decision-making. By predicting anomalies, the system can take measures in advance to prevent the problem from further expanding;

[0156] Through drone inspections and predictive analysis, the system can identify equipment that requires maintenance and prevent faults from expanding. When a fault is detected, the system can automatically isolate the faulty part to protect other equipment from being affected, ensuring the overall safety of the power plant. When a problem is detected, the system automatically adjusts operating parameters to reduce the impact on power plant operations, and achieves self-healing through continuous monitoring to prevent small problems from developing into major faults.

[0157] Embodiment 2:

[0158] Please refer to Figures 3 to 6 As shown, a new energy power station sensing control system includes:

[0159] The data processing and analysis module includes a data acquisition and preprocessing unit and a data analysis and anomaly detection unit. It performs denoising and standardization on the collected sensor data, analyzes and predicts the operating status of the new energy power station by training a deep neural network model, identifies abnormal situations, including power fluctuations and voltage anomalies, and calculates the health index of the equipment to evaluate the current status and expected life of the equipment.

[0160] The control and decision-making module includes an energy management unit and a multi-level response strategy unit. The energy management unit dynamically adjusts the energy production and distribution strategy of the power station according to real-time data, optimizes the operation of photovoltaic panels and energy storage systems, and adjusts the wind turbine angle, photovoltaic panel angle and working state of the energy storage system in real time to maximize energy output and extend equipment life. The multi-level response strategy unit includes low-level response, medium-level response and high-level response;

[0161] Intelligent operation and maintenance module, which predicts potential equipment failures based on historical data and deep learning models, plans maintenance and repair work in advance to avoid sudden failures, and uses drones to regularly inspect photovoltaic panels and other key equipment, capture image and video data, and identify damage or contamination;

[0162] The feedback and self-healing module includes a closed-loop feedback control unit and a self-healing function unit. It provides real-time feedback on the execution effect of the control strategy and dynamically adjusts the control plan based on the feedback data to ensure that the power station is always in the best operating state. When equipment failure or operation deviation is detected, the system automatically adjusts the operating parameters, implements the self-healing function, restores the equipment to normal state, and reduces downtime and operation risks.

[0163] Specifically, low-level response: automatically adjust operating parameters to respond to minor anomalies to ensure continued stable operation of the power plant;

[0164] Medium-level response: When a moderate anomaly is detected, the system automatically adjusts the equipment and notifies the operation and maintenance personnel to conduct an inspection.

[0165] High-level response: For serious anomalies, the system performs an emergency shutdown and issues an alarm, requiring human intervention to prevent the accident from spreading.

[0166] As can be seen above, the system adopts a hierarchical response plan based on the severity of the anomaly, from automatic adjustment to manual intervention. Low-level anomalies can be automatically repaired by the system to avoid delays in manual intervention; high-level anomalies are promptly notified to the operation and maintenance personnel for processing to prevent serious failures.

[0167] By adjusting the photovoltaic array angle and the charging and discharging strategy of the energy storage system in real time, the system can effectively reduce power generation losses caused by power fluctuations or voltage anomalies. The automatic adjustment function is particularly prominent in low-level anomalies, keeping the power station in the best operating state.

[0168] Embodiment three:

[0169] This design is specifically applied in solar photovoltaic power stations. It uses the powerful learning ability of deep neural networks (DNNs) to deeply mine environmental parameters, photovoltaic panel output power, battery energy storage status and grid interaction data collected in real time by massive sensors. By continuously analyzing subtle changes in these data streams, the system can accurately identify tiny signs of power fluctuations. For minor or common abnormalities, the system can automatically make fine adjustments or repairs without human intervention.

[0170] Furthermore, a photovoltaic power station is located in a remote plateau area with good lighting conditions, but changeable climate conditions, often accompanied by strong winds, low temperatures and humidity changes. The power station integrates energy storage systems and wind power generation equipment to ensure stable power supply under different weather conditions. In order to further improve the operating efficiency of the power station and reduce power generation losses and shortened equipment life caused by abnormal conditions, the power station decided to deploy a new energy power station perception control system and implement a multi-level response strategy, as follows:

[0171] Data collection:

[0172] Deploy a multi-source sensor network:

[0173] Light sensors: Arranged on photovoltaic panels at different angles, each sensor collects light data once a minute.

[0174] Data volume: 60 light sensors, collecting data 1440 times per day.

[0175] Data format: Light intensity (W / m 2 ).

[0176] Temperature sensors: Distributed inside and outside the power station, used to monitor ambient temperature and equipment temperature.

[0177] Data volume: 20 temperature sensors, collecting data 1440 times per day.

[0178] Data format: Temperature (℃).

[0179] Wind speed sensor: placed near the wind turbine to monitor wind speed changes.

[0180] Data volume: 10 wind speed sensors, collecting data 1440 times per day.

[0181] Data format: wind speed (m / s).

[0182] Voltage sensor: connected to each major power output node of the power station to monitor the voltage level in real time.

[0183] Data volume: 30 voltage sensors, collecting data 1440 times per day.

[0184] Data format: Voltage (V).

[0185] Current sensor: installed at the output of photovoltaic panels and wind turbines to monitor current flow.

[0186] Data volume: 30 current sensors, collecting data 1440 times per day.

[0187] Data format: Current (A).

[0188] Data processing and storage:

[0189] Data collection: Transmitted to the central processor in real time through wireless sensor networks.

[0190] Data cleaning and preprocessing: During the processing, all sensor data are cleaned and filtered outliers every day, and standardized.

[0191] Data volume: approximately 2 million data points per day.

[0192] Data storage: Daily data is approximately 2GB, stored in a local database and uploaded to the cloud weekly for long-term preservation and in-depth analysis.

[0193] State Awareness:

[0194] Monitor the collected data in real time and use the Adaptive Health Index (AHI) to assess the health status of the equipment;

[0195] Monitoring and early warning:

[0196] Design and training of deep learning models:

[0197] Model input:

[0198] Multi-dimensional input: light intensity, temperature, wind speed, voltage, and current data.

[0199] Data Window: Each input data window contains the sensor data for the past 10 minutes (1440×5 dimensional data).

[0200] Model structure:

[0201] Deep Neural Networks (DNN):

[0202] Input layer: receives all sensor data within the past 10 minutes.

[0203] Hidden layer: 3 layers, each layer contains 512 neurons, and the activation function is ReLU.

[0204] Output layer: multi-classification output, including normal, low-level abnormality, medium-level abnormality, and high-level abnormality.

[0205] Model training:

[0206] Data volume: The model is trained using historical sensor data from the past year (approximately 700 GB).

[0207] Training process: supervised learning is used to calibrate the model using labeled abnormal event data. Cross-validation and hyperparameter optimization are used during the training process to improve the generalization ability of the model.

[0208] Model Accuracy: The final model achieved 95% accuracy and 90% recall on the test set.

[0209] Anomaly detection and response strategy

[0210] Real-time monitoring:

[0211] The model runs once a minute, analyzing the current operating status of the power plant in real time and predicting whether there are any abnormalities.

[0212] The output results include: no abnormality, slight abnormality, moderate abnormality, and severe abnormality.

[0213] Multi-level response strategy:

[0214] A multi-level response strategy is used to automatically select the most appropriate response measure according to the severity of the anomaly, and its expression is:

[0215]

[0216] in The response strategy that the system should adopt for the currently detected anomaly;

[0217] P(c1|X): The probability of a low-level anomaly (such as a slight power fluctuation).

[0218] P(c2|X): probability of medium-level anomaly (such as moderate voltage anomaly);

[0219] P(c3|X): probability of high-level anomalies (such as severe voltage anomalies);

[0220] θ1, θ2: probability thresholds, 0.7 and 0.4 respectively;

[0221] Minor abnormalities:

[0222] Response measures: Automatically adjust the angle of photovoltaic panels and optimize the charging and discharging strategy of the energy storage system.

[0223] Processing time: The system response time is about 5 seconds, and the abnormality elimination time is about 2 minutes.

[0224] The light sensor detects light fluctuations during a certain period of time, and the temperature sensor shows that the device temperature is slightly high. The system automatically adjusts the angle of the photovoltaic panel to reduce the temperature impact.

[0225] Moderate abnormality:

[0226] Response measures: The system automatically adjusts parameters and sends a notification to the operation and maintenance personnel, suggesting an on-site inspection.

[0227] Processing time: The system response time is about 10 seconds, and the abnormality is usually eliminated within 30 minutes.

[0228] The wind speed sensor shows a sudden increase in wind speed, the voltage sensor detects fluctuations, the system adjusts the fan operating status and notifies the operation and maintenance personnel to check.

[0229] Severe abnormalities:

[0230] Response measures: The system shuts down immediately and issues an emergency alarm, requiring human intervention.

[0231] Processing time: The system response time is about 1 second, and the abnormality elimination time depends on the on-site repair situation.

[0232] Voltage and current sensors detect serious anomalies that could cause equipment damage, and the system immediately triggers a shutdown procedure.

[0233] The system continuously monitors the effectiveness of control measures and adjusts control strategies based on feedback data. If the abnormal situation persists or worsens, the response level will be increased.

[0234] Intelligent Operation and Maintenance:

[0235] Intelligent operation and maintenance platform: Through the drone inspection system, the status of photovoltaic panels is regularly checked and maintenance plans are automatically generated.

[0236] Based on the deep learning prediction model, equipment that needs maintenance can be identified in advance to avoid sudden failures.

[0237] Self-healing dispatch: When the system detects a local fault, it automatically isolates the faulty part to prevent the impact from spreading to the entire power station. The system automatically adjusts operating parameters based on the prediction results to reduce the impact of the fault.

[0238] The result is as follows:

[0239] Improved power generation efficiency: Through automated regulation and optimization, power generation is increased by approximately 3% each year, reducing power generation losses due to abnormal conditions.

[0240] Extended equipment life: By implementing accurate predictive maintenance, the average service life of equipment has been extended by about 15%, greatly reducing maintenance costs.

[0241] The anomaly detection accuracy rate reaches over 95%, especially the recognition accuracy of power fluctuations is high.

[0242] The automatic regulation of low-level abnormalities is 90% effective, reducing power loss by 20%.

[0243] Through a multi-level response strategy, downtime for high-level exceptions was reduced by 30%.

[0244] Improved abnormal response efficiency: The implementation of the multi-level response strategy has reduced the reaction time of operation and maintenance personnel by about 50% and shortened the abnormality elimination time by about 30%.

[0245] When a medium-level abnormality occurs, the system can notify the operation and maintenance personnel within 5 minutes and automatically adjust relevant parameters.

[0246] After implementing the self-healing feature, downtime caused by severe anomalies was reduced by 40% through early detection and automated response.

[0247] From the above, we can see that through accurate abnormal monitoring and the implementation of multi-level response strategies, the power station not only significantly improved the operating efficiency, but also greatly extended the service life of the equipment, ultimately achieving higher economic benefits and safety.

[0248] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A new energy power station sensing control method, characterized in that: include: S1. Data collection: arrange multi-source sensor networks, collect parameters of new energy equipment, deploy edge computing nodes, preliminarily process and filter the data collected by sensors, reduce redundant data transmission, obtain external data such as weather, load forecast, and market electricity prices in real time, and integrate them with historical data to form comprehensive data input; S2. Status perception: real-time monitoring of the operating status of power generation units, energy storage systems, and transmission and distribution network subsystems to ensure comprehensive data coverage. The adaptive health index method is used to evaluate the health status of equipment, predict the probability of failure, and estimate the remaining service life of equipment through big data analysis and machine learning algorithms. S3, monitoring and early warning, using deep learning algorithms to monitor the operation status of the power station in real time, identify power fluctuations and voltage anomalies, design multi-level response strategies, and formulate different response plans according to the severity of the anomaly, including automatic adjustment for low-level anomalies, automatic adjustment for medium-level anomalies and notification of operation and maintenance, and manual intervention for high-level anomalies; S4, decision-making and optimization control, using intelligent energy scheduling, combined with real-time load demand, energy storage status, and weather forecast factors to optimize energy distribution and use, and using reinforcement learning algorithms to optimize power station operating parameters in real time, including power factor and conversion efficiency; S5, Execution and feedback. Each subsystem in the power station has autonomous control capabilities. When the central system fails, each subsystem can work together to maintain the basic operating status of the system, establish a closed-loop feedback mechanism, continuously monitor the execution effect of control instructions, and feed back the results to the decision-making layer. The control strategy is automatically adjusted according to the feedback data to improve the system's response speed and decision-making accuracy; S6. Intelligent operation and maintenance: Develop an intelligent operation and maintenance platform that can automatically generate maintenance plans, and use remote operation and maintenance robots to detect and repair equipment. Use drone inspections to regularly inspect the surface cleanliness of photovoltaic panels and the wear of wind turbine blades to ensure that the equipment is in the best condition. S7, self-healing scheduling, when a local fault is detected, the fault area is automatically isolated to prevent the impact from spreading to the entire power station. When a potential problem is detected, the system automatically adjusts the operating strategy until the fault is completely repaired.

2. A new energy power station sensing control method according to claim 1, characterized in that: The multi-source sensor network includes photovoltaic power station sensors, wind power generation sensors, energy storage system sensors, environmental monitoring sensors, grid access monitoring sensors and safety protection sensors.

3. A new energy power station sensing control method according to claim 1, characterized in that: The comprehensive data in step S1 is obtained through data fusion, and its expression is: Dt=α·Dr+β·Dh+γ·Df Where Dt: the fused data set, used for subsequent analysis and decision-making; Dr: Data collected in real time; Dh: historical data, reflecting the past behavior and trends of the system; Df: forecast data, including weather forecast and load forecast; α, β, γ: weight coefficients, reflecting the contribution of each type of data to the final fusion result, and satisfying α+β+γ=1.

4. A new energy power station sensing control method according to claim 1, characterized in that: In step S2, the adaptive health index AHi calculates the current health status of the device by standardizing various monitoring data, and can detect potential problems in time. The expression is: AHi: The health index of a device or system. A higher value indicates a worse health status. Xj: jth monitoring data, including temperature and voltage; μj: the historical average value of the j-th data; σj: standard deviation of the j-th data; n: the number of monitoring data.

5. A new energy power station sensing control method according to claim 1, characterized in that: In step S3, the deep learning algorithm uses a multi-level response strategy to automatically select the most appropriate response measure according to the severity of the abnormality, and its expression is: in The response strategy that the system should adopt for the currently detected anomaly; P(c1|X): probability of low-level anomaly; P(c2|X): probability of medium-level anomaly; P(c3|X): probability of high-level anomaly; θ1, θ2: probability thresholds that determine the switching point of the response strategy.

6. A new energy power station sensing control method according to claim 1, characterized in that: The intelligent energy scheduling in step S4 is expressed as: Po: optimal scheduling power allocation; Pi: power output of the i-th energy unit; C(Pi): the power cost function of the i-th energy unit, which is usually related to the power generation cost and energy consumption; R(Pi): Power volatility or risk index, measuring the impact of uncertainty on the system; λ: Risk preference coefficient, reflecting the system's tolerance to volatility risk; m: the number of energy units.

7. A new energy power station sensing control method according to claim 1, characterized in that: In step S4, the reinforcement learning algorithm gradually approaches the optimal strategy by continuously updating the value function Q(s,a) to achieve dynamic optimization control, specifically: Where Q(s,a): the value function of taking action a in state s; α: learning rate, which determines the update step size; r: immediate reward, reflecting the direct benefit of the current action; γ: discount factor, measuring the current value of future earnings; s′: the new state the system transfers to after executing action a; a′: action in the new state s′.

8. A new energy power station sensing control method according to claim 1, characterized in that: The expression for isolating the fault in step S7 is: Where I(t): fault isolation index, used to evaluate whether a fault area needs to be isolated; Fk(t): the value of the kth fault indicator at time t; wk: weight of the kth fault indicator, reflecting its importance; K: Number of fault indicators.

9. A new energy power station sensing control system, characterized in that: include: The data processing and analysis module includes a data acquisition and preprocessing unit and a data analysis and anomaly detection unit. It performs denoising and standardization on the collected sensor data, analyzes and predicts the operating status of the new energy power station by training a deep neural network model, identifies abnormal situations, including power fluctuations and voltage anomalies, and calculates the health index of the equipment to evaluate the current status and expected life of the equipment. The control and decision-making module includes an energy management unit and a multi-level response strategy unit. The energy management unit dynamically adjusts the energy production and distribution strategy of the power station according to real-time data, optimizes the operation of photovoltaic panels and energy storage systems, and adjusts the wind turbine angle, photovoltaic panel angle and working state of the energy storage system in real time to maximize energy output and extend equipment life. The multi-level response strategy unit includes low-level response, medium-level response and high-level response; Intelligent operation and maintenance module, which predicts potential equipment failures based on historical data and deep learning models, plans maintenance and repair work in advance to avoid sudden failures, and uses drones to regularly inspect photovoltaic panels and other key equipment, capture image and video data, and identify damage or contamination; The feedback and self-healing module includes a closed-loop feedback control unit and a self-healing function unit. It provides real-time feedback on the execution effect of the control strategy and dynamically adjusts the control plan based on the feedback data to ensure that the power station is always in the best operating state. When equipment failure or operation deviation is detected, the system automatically adjusts the operating parameters, implements the self-healing function, restores the equipment to normal state, and reduces downtime and operation risks.

10. A new energy power station sensing control system according to claim 9, characterized in that: The low-level response: automatically adjusts operating parameters to deal with minor anomalies to ensure continuous and stable operation of the power plant; Medium-level response: When a moderate anomaly is detected, the system automatically adjusts the equipment and notifies the operation and maintenance personnel to conduct an inspection. High-level response: For serious anomalies, the system performs an emergency shutdown and issues an alarm, requiring human intervention to prevent the accident from spreading.

Images