Hydroelectric power station monitoring system and control method

Abstract

The invention discloses a hydroelectric power station monitoring system and a control method, and relates to the technical field of intelligent monitoring, and the system comprises the steps: carrying out the calibration of a plurality of sensors, collecting the data, uploading the data of the plurality of sensors to a cloud end in real time through an Internet of Things LoRaWAN network, and setting a data receiving and storing unit at the cloud end; preprocessing the received original data of different sensors, and fusing the preprocessed data by adopting a simple averaging method; calculating the actual energy output of the hydroelectric power station based on the comprehensive evaluation report, and performing energy optimization through a reinforcement learning model; and executing a control strategy according to an optimization result to realize optimal energy output of the hydroelectric power station. Through the DQN reinforcement learning model, the actual energy output is calculated according to the real-time sensor data and compared with the target energy output, the reward value is generated, self-adaptive control is achieved, dynamic adjustment can be achieved in the complex and changeable environment, and the optimal energy output is achieved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent monitoring technology, in particular to a monitoring system and a control method for a hydropower station. Background Art

[0002] In recent years, with the development of renewable energy technology, hydropower generation, as a clean and sustainable energy source, has been widely used around the world; the traditional monitoring system of hydropower stations mainly relies on manual inspections and simple sensor data collection, and ensures the normal operation of the power station by regularly checking the equipment status and operating parameters; in recent years, with the development of Internet of Things technology, big data processing and artificial intelligence, the monitoring and control methods of hydropower stations are gradually moving towards intelligence; modern hydropower station monitoring systems integrate a variety of sensors to collect key parameters such as water level, water flow rate, temperature, pressure, etc. in real time, and use cloud computing platforms for data storage and analysis, thereby realizing comprehensive monitoring and intelligent control of the operating status of the power station.

[0003] There are still many problems in the existing monitoring systems of hydropower stations. First, data collection and processing in traditional systems often lack an efficient data fusion mechanism, resulting in the failure to fully utilize data obtained from different sensors. Second, existing data analysis methods are mostly based on static models, which are difficult to cope with complex and changeable environmental conditions and real-time needs, limiting prediction accuracy and response speed. In addition, the control strategies of existing systems are usually simple and lack adaptive adjustment capabilities, making it impossible to dynamically optimize energy output according to actual operating conditions. The present invention proposes a monitoring system and control method for a hydropower station, which effectively solves the above problems by introducing advanced data fusion technology, reinforcement learning algorithms and adaptive control strategies, and achieves more accurate energy output prediction and optimization control, significantly improving the operating efficiency and reliability of the hydropower station. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a monitoring system and control method for a hydropower station to solve the problems of energy output optimization and real-time control of a hydropower station.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides a method for monitoring and controlling a hydropower station, which includes calibrating and collecting data from multiple sensors, uploading the data from multiple sensors to the cloud in real time through the Internet of Things LoRaWAN network, and setting a data receiving and storage unit on the cloud; preprocessing the raw data received from different sensors, and fusing the preprocessed data using a simple averaging method; analyzing the historical data and real-time data of the fused sensor data using the Hadoop big data processing framework to generate a comprehensive evaluation report on the operating status of the hydropower station; based on the comprehensive evaluation report, calculating the actual energy output of the hydropower station, and optimizing the energy through a reinforcement learning model; and executing a control strategy according to the optimization result to achieve the optimal energy output of the hydropower station.

[0008] As a preferred solution of the hydropower station monitoring and control method of the present invention, the multiple sensors are calibrated and data is collected, the multiple sensor data are uploaded to the cloud in real time through the Internet of Things LoRaWAN network, and a data receiving and storage unit is set in the cloud. The specific steps are:

[0009] Calibrate water machine flow meters, water level sensors, pressure sensors and temperature sensors;

[0010] Upload calibrated sensor data to the cloud in real time via the LoRaWAN network;

[0011] Create an IoT core in the cloud using AWS IoT Core;

[0012] Choose Amazon DynamoDB database as the repository for structured data;

[0013] Create a DynamoDB table in the AWS console and set the read and write capacity units;

[0014] Define the primary key device string type, timestamp numeric type, data numeric type, status string type, and location string type in the DynamoDB table;

[0015] Convert sensor data into a unified JSON format and store it in an Amazon DynamoDB database;

[0016] Use AWS SDK to store the received data in a DynamoDB table;

[0017] Create a global secondary GSI index for frequently accessed fields;

[0018] Generate unique certificates and keys for sensor devices;

[0019] Enable the static encryption feature of DynamoDB to encrypt the stored data using the KMS key;

[0020] Use IAM policies to strictly control access to DynamoDB tables;

[0021] Configure the AWS IoT rule engine to send SNS notifications when water level, flow, temperature, and device status change.

[0022] As a preferred solution of the hydropower station monitoring and control method of the present invention, the raw data received from different sensors are preprocessed, and the data from multiple sensors are fused using a simple averaging method. The specific steps are:

[0023] The IQR interquartile range method is used to detect and remove data outliers from the water machine flow meter data, the spline interpolation method is used to fill the missing values ​​of the data, and the Savitzky-Golay filter is used to smooth the data to remove the noise of the water machine flow meter data;

[0024] The Z-Score statistical method is used to detect and remove data outliers from the water level sensor data, the linear interpolation method is used to fill the missing values ​​of the data, and the five-point moving average method is used to smooth the data to remove the noise of the water level sensor data;

[0025] The local outlier factor (LOF) algorithm is used to detect and remove data outliers from the pressure sensor data. The missing values ​​of the data are filled by the Lagrange interpolation method. The median filter is used to smooth the data and remove the noise of the pressure sensor data.

[0026] The DBSCAN clustering algorithm is used to detect and remove data outliers from the temperature sensor data, the polynomial interpolation method is used to fill the missing values ​​of the data, and the EWMA exponentially weighted moving average method is used to smooth the data and remove the noise of the temperature sensor data;

[0027] All preprocessed sensor data are fused through a simple averaging method to obtain fused sensor data.

[0028] As a preferred solution of the monitoring and control method of the hydropower station of the present invention, the fused sensor data is analyzed by using the Hadoop big data processing framework to analyze the historical data and real-time data to generate a comprehensive evaluation report on the operation status of the hydropower station. The specific steps are:

[0029] The collected real-time data is processed in real time using Apache Kafka and Apache Flink, tools in the Hadoop big data processing framework, to generate an evaluation report on the power plant operation status of real-time data, and to display the real-time data and analysis results through charts and dashboards;

[0030] Extract historical data from the cloud database, use Hadoop MapReduce to perform batch analysis on the historical data, and use the ARIMA model combined with the LSTM neural network to predict future water level trends;

[0031] Combine real-time data with historical data analysis results to generate a comprehensive assessment report on the power plant's operating status.

[0032] As a preferred solution of the method for monitoring and controlling a hydropower station of the present invention, the actual energy output of the hydropower station is calculated based on the comprehensive evaluation report, and energy optimization is performed through a reinforcement learning model. The specific steps are:

[0033] Extract the current water level data H(t) and water flow velocity data V(t) from the comprehensive evaluation report, and calculate the actual energy output P(t) at the current time. The expression is:

[0034] P(t)=αH(t)+βV(t);

[0035] Among them, t is the time point, α is the empirical coefficient of the current water level, and β is the empirical coefficient of the water flow velocity;

[0036] Define the current water level data and water flow speed data as the current state;

[0037] Based on the actual energy output calculated from the current water level data and water flow velocity, the initial blade angle and water flow channel opening size are generated;

[0038] Define the initial blade angle as θ i (t) and the initial water flow channel opening size is S i (t) as a regulatory action;

[0039] The target energy output value is determined based on the power station design capacity, the real-time demand of the power grid dispatching center, the environmental conditions of the current water level data and water flow velocity data, and the economic costs and benefits;

[0040] According to the difference between the actual energy output and the target energy output, the initial reward is calculated as follows:

[0041]

[0042] Among them, r init is the initial reward value, Ptg is the target energy output value.

[0043] As a preferred embodiment of the method for monitoring and controlling a hydropower station according to the present invention, the energy optimization by using a reinforcement learning model refers to:

[0044] Select the Deep Q-Network reinforcement learning model;

[0045] Collect historical data and real-time data to build training data sets;

[0046] Initialize Deep Q-Network model parameters;

[0047] The Deep Q-Network model selects and adjusts actions based on the current state;

[0048] Perform adjustment actions and observe the state and actual energy output at the next time;

[0049] According to the next actual energy output and target energy output, the optimized reward is calculated as follows:

[0050]

[0051] Where P(t+1) is the energy output value actually measured at time t+1;

[0052] Store the current state, action, reward, and next state in the experience replay buffer in the Deep Q-Network model;

[0053] Randomly sample a batch of experience data from the experience replay buffer and update the Q value of the Deep Q-Network reinforcement learning model;

[0054] The Deep Q-Network model achieves optimization by continuously learning to make the actual energy output equal to the target energy output.

[0055] As a preferred solution of the monitoring and control method of the hydropower station of the present invention, the control strategy is executed according to the optimization result to achieve the optimal energy output of the hydropower station. The specific steps are:

[0056] According to the optimized energy, the angle θ(t) of the turbine blade is adjusted, and the expression is:

[0057] θ(t)=γP(t)+δ;

[0058] Among them, γ is the influence coefficient of energy output on blade angle, and δ is the setting value of blade angle;

[0059] According to the water flow velocity and blade angle, adjust the water flow channel opening size S(t), the expression is:

[0060] S(t)=ηV(t)+ζθ(t);

[0061] Among them, η is the empirical coefficient of the influence of water flow velocity on the opening size of the water flow channel, and ζ is the empirical coefficient of the influence of blade angle on the opening size of the water flow channel.

[0062] In a second aspect, the present invention provides a monitoring and control system for a hydropower station, comprising a calibration module, a processing module, a data analysis module, an optimization module and a control execution module; the calibration module is used to calibrate a variety of sensors and collect data, upload the data of a variety of sensors to the cloud in real time through the Internet of Things LoRaWAN network, and set a data receiving and storage unit on the cloud; the processing module is used to pre-process the raw data received from different sensors, and fuse the pre-processed data using a simple averaging method; the data analysis module is used to analyze the historical data and real-time data of the fused sensor data using the Hadoop big data processing framework, and generate a comprehensive evaluation report on the operating status of the hydropower station; the optimization module is used to calculate the actual energy output of the hydropower station based on the comprehensive evaluation report, and optimize the energy through a reinforcement learning model; the control execution module is used to execute the control strategy according to the optimization result to achieve the optimal energy output of the hydropower station.

[0063] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the hydropower station monitoring and control method as described in the first aspect of the present invention is implemented.

[0064] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the method for monitoring and controlling a hydropower station as described in the first aspect of the present invention is implemented.

[0065] The beneficial effects of the present invention are as follows: the present invention uses the DQN reinforcement learning model to calculate the actual energy output according to the real-time sensor data and compare it with the target energy output to generate a reward value. The DQN model selects the optimal action according to the current state and gradually optimizes the control strategy by continuously updating the Q value. This method not only realizes adaptive control, but also can dynamically adjust in a complex and changing environment to achieve the optimal energy output. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0067] Figure 1 This is a flow chart of the monitoring and control method for a hydropower station in Example 1.

[0068] Figure 2 This is a flow chart of using Hadoop to analyze data and produce a comprehensive evaluation report in Example 1. DETAILED DESCRIPTION

[0069] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0070] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0071] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0072] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and provides a method for monitoring and controlling a hydropower station, comprising the following steps: calibrating a plurality of sensors and collecting data, uploading the plurality of sensor data to the cloud in real time through the Internet of Things LoRaWAN network, and setting a data receiving and storage unit in the cloud; preprocessing the raw data received from different sensors, and fusing the preprocessed data using a simple averaging method; analyzing the historical data and real-time data of the fused sensor data using the Hadoop big data processing framework, and generating a comprehensive evaluation report on the operating status of the hydropower station; based on the comprehensive evaluation report, calculating the actual energy output of the hydropower station, and optimizing the energy through a reinforcement learning model; executing a control strategy according to the optimization result to achieve the optimal energy output of the hydropower station.

[0073] S1. Calibrate various sensors and collect data, upload various sensor data to the cloud in real time through the IoT LoRaWAN network, and set up data receiving and storage units in the cloud.

[0074] Calibrate water machine flow meter, water level sensor, pressure sensor, temperature sensor;

[0075] Upload calibrated sensor data to the cloud in real time via the LoRaWAN network;

[0076] Use AWS IoT Core in the cloud to create an IoT core as a bridge between devices and the cloud;

[0077] Choose Amazon DynamoDB database as the repository for structured data;

[0078] Create a DynamoDB table in the AWS console and set the read and write capacity units;

[0079] Define the primary key device string type and timestamp numeric type, data numeric type, status string type and location string type in the DynamoDB table;

[0080] Convert sensor data into a unified JSON format and store it in an Amazon DynamoDB database;

[0081] Use AWS SDK to store the received data in a DynamoDB table;

[0082] Create a global secondary GSI index for frequently accessed fields to improve query efficiency;

[0083] Generate unique certificates and keys for sensor devices to ensure secure data transmission;

[0084] Enable the static encryption feature of DynamoDB to encrypt the stored data using the KMS key;

[0085] Use IAM policies to strictly control access to DynamoDB tables;

[0086] Configure the AWS IoT rule engine to send SNS notifications when water level, flow, temperature, and device status change.

[0087] It should be noted that this step calibrates multiple sensors and uploads data to the cloud in real time, performs preprocessing to remove outliers, fill missing values, and smooth data, and finally uses a simple averaging method to fuse the data to ensure the accuracy and reliability of the data, providing a high-quality data foundation for subsequent analysis.

[0088] S1.1. Create a DynamoDB table in the AWS console.

[0089] Define the primary key device ID as DeviceID, and the type is String;

[0090] The timestamp is Timestamp, and the type is Number;

[0091] Define the water level data of the field as WaterLevel, and the type is Number;

[0092] The flow data type is FlowRate, and the type is Number;

[0093] The pressure data type is Pressure, and the type is Number;

[0094] The temperature data type is Temperature, and the type is Number;

[0095] The status is Status, and the type is String;

[0096] The location is Location, and the type is String;

[0097] DeviceID is the unique identifier of the device, used to distinguish different sensor devices;

[0098] Timestamp is the timestamp of recorded data, which is used to store and query data in chronological order;

[0099] WaterLevel is the current water level data, expressed as a floating point number;

[0100] FlowRate is the current water flow rate data, expressed as a floating point number;

[0101] Pressure is the current pressure data, expressed as a floating point number;

[0102] Temperature is the current temperature data, expressed as a floating point number;

[0103] Status is the status information of the device, such as "normal", "fault", etc., expressed as a string;

[0104] Location is the location information of the device, expressed as a string.

[0105] S2. Preprocess the raw data received from different sensors and fuse the data from multiple sensors using a simple averaging method.

[0106] The IQR interquartile range method is used to detect and remove data outliers from the water machine flow meter data, the spline interpolation method is used to fill the missing values ​​of the data, and the Savitzky-Golay filter is used to smooth the data to remove the noise of the water machine flow meter data;

[0107] The Z-Score statistical method is used to detect and remove data outliers from the water level sensor data, the linear interpolation method is used to fill the missing values ​​of the data, and the five-point moving average method is used to smooth the data to remove the noise of the water level sensor data;

[0108] The local outlier factor (LOF) algorithm is used to detect and remove data outliers from the pressure sensor data. The missing values ​​of the data are filled by the Lagrange interpolation method. The median filter is used to smooth the data and remove the noise of the pressure sensor data.

[0109] The DBSCAN clustering algorithm is used to detect and remove data outliers from the temperature sensor data, the polynomial interpolation method is used to fill the missing values ​​of the data, and the EWMA exponentially weighted moving average method is used to smooth the data and remove the noise of the temperature sensor data;

[0110] All preprocessed sensor data are fused through a simple averaging method to obtain fused sensor data.

[0111] It should be noted that this step improves the quality of multiple sensor data through preprocessing methods such as outlier detection, missing value filling and data smoothing, and uses a simple averaging method to fuse the data to ensure the accuracy and consistency of the final data, providing a reliable basis for subsequent analysis.

[0112] S2.1. Further explanation:

[0113] The preprocessing of water flow meter data is as follows:

[0114] The IQR interquartile range method was used to detect and remove data outliers;

[0115] Calculate the first and third quartiles;

[0116] The interquartile range was calculated and expressed as: IQR = Q3-Q1;

[0117] The outlier range was defined as: [Q1-1.5×IQR, Q3+1.5×IQR];

[0118] Data outside the defined outlier range are considered outliers and removed;

[0119] Use spline interpolation to fill in missing values ​​in the data. Spline interpolation is a smooth interpolation method suitable for continuous data.

[0120] Use Savitzky-Golay filter to reduce data noise. Savitzky-Golay filter is a smoothing filter based on polynomial fitting, which can effectively remove high-frequency noise.

[0121] The preprocessing of water level sensor data is as follows:

[0122] Calculate the Z-Score of each data point, the expression is:

[0123] Where X is the data point, μ is the mean, and σ is the standard deviation;

[0124] Set a threshold, for example, set it to 3. If the data exceeds the threshold, it is considered an outlier and the outlier data is removed.

[0125] Fill missing values ​​by linear interpolation, which is a simple interpolation method that assumes that the data varies linearly between two known points;

[0126] Use the five-point moving average method to smooth the data and reduce short-term fluctuations. The five-point moving average method calculates the average of five consecutive data points to smooth the data;

[0127] The preprocessing of pressure sensor data is as follows:

[0128] Use the local outlier factor (LOF) algorithm to detect and remove outliers. The LOF algorithm identifies local outliers by comparing the density of data points with their domain.

[0129] Fill missing values ​​through Lagrange interpolation, which is a polynomial interpolation method suitable for non-uniformly distributed data;

[0130] Use a median filter to smooth data and reduce burst noise. The median filter smoothes data by taking the median value within a certain window, which is particularly suitable for removing impulse noise;

[0131] The preprocessing of temperature sensor data is as follows:

[0132] Use the DBSCAN clustering algorithm to detect and remove outliers. The DBSCAN algorithm identifies outliers through density clustering.

[0133] Use polynomial interpolation to fill missing values. Polynomial interpolation is to fill missing values ​​by polynomial fitting;

[0134] Use EWMA exponential weighted moving average method to smooth the data and reduce data noise;

[0135] The EWMA formula is: t =λZ t+(1-λ)Y t -1;

[0136] Among them, Y t is the smoothed value at time t, Z t is the actual observation value at time t, λ is the smoothing coefficient, which indicates the influence of the current observation value on the smoothed value, and Y t -1 is the smoothing value of the previous time point t-1; after the data preprocessing is completed, all preprocessed sensor data are fused by simple averaging method;

[0137] The formula is:

[0138] Among them, Fused Data is the average value of fusion, D i is the data of the i-th sensor, n is the number of sensors;

[0139] S2.1.1 Example:

[0140] Assume that the data of four sensors: water machine flow meter, water level sensor, pressure sensor, and temperature sensor are:

[0141] Water machine flow meter data: 10.2, 10.5, 10.3, 10.4, 10.6;

[0142] Water level sensor data: 5.1, 5.2, 5.3, 5.2, 5.1;

[0143] Pressure sensor data: 2.1, 2.0, 2.2, 2.1, 2.0;

[0144] Temperature sensor data: 25.1, 25.2, 25.3, 25.2, 25.1;

[0145] The average value of each group is obtained by the formula:

[0146] The average value of the water machine flow meter data is:

[0147] The average value of the water level sensor data is:

[0148] The average value of the pressure sensor data is:

[0149] The average value of the temperature sensor data is:

[0150] The average values ​​of the sensors calculated above are combined by simple averaging method and expressed as follows:

[0151]

[0152] Finally, the data from multiple sensors are fused through a simple averaging method to obtain a comprehensive value, providing a reliable data basis for subsequent big data analysis and energy output optimization.

[0153] S3. Use the Hadoop big data processing framework to analyze the historical data and real-time data after the fusion, and generate a comprehensive evaluation report on the operating status of the hydropower station.

[0154] The cloud receives data from a variety of sensors;

[0155] The received data is stored in the DynamoDB cloud database, and the time information of each data is recorded using a timestamp;

[0156] Real-time data and historical data are tested using the Z-Score statistical method, outliers are removed, and missing values ​​are filled using spline interpolation;

[0157] The collected real-time data is processed in real time using Apache Kafka and Apache Flink, tools in the Hadoop big data processing framework, to generate an evaluation report on the power plant operation status based on real-time data. The real-time data and analysis results are displayed through charts and dashboards, so that operators can intuitively understand the power plant operation status.

[0158] Extract historical data from the cloud database, use Hadoop MapReduce to perform batch analysis on the historical data, and use the ARIMA model combined with the LSTM neural network to predict future water level trends;

[0159] Combine real-time data with historical data analysis results to generate a comprehensive assessment report on the power plant's operating status;

[0160] If an abnormal situation is detected, an alarm will be automatically triggered and relevant technicians will be notified via mobile phone.

[0161] It should be noted that this step processes the sensor data received from the cloud through the Hadoop big data framework, performs real-time and historical data analysis, generates a comprehensive evaluation report on the power plant's operating status, and ensures efficient and stable operation of the power plant through visual displays and automatic alarm systems.

[0162] S3.1. The report content includes: current water level and water flow rate, equipment status, predicted future water level change trend, abnormal conditions and potential risks;

[0163] The alarm information includes: abnormal parameters and their values, timestamp, and recommended measures.

[0164] S4. Based on the comprehensive evaluation report, the actual energy output of the hydropower station is calculated, and the difference between the actual energy output and the target energy output is optimized through the reinforcement learning model.

[0165] Extract the current water level data H(t) and water flow velocity data V(t) from the comprehensive evaluation report, and calculate the actual energy output P(t) at the current time. The expression is:

[0166] P(t)=αH(t)+βV(t);

[0167] Among them, t is the time point, α is the empirical coefficient of the current water level, and β is the empirical coefficient of the water flow velocity;

[0168] Define the current water level data and water flow speed data as the current state;

[0169] Based on the actual energy output calculated from the current water level data and water flow velocity, the initial blade angle and water flow channel opening size are generated;

[0170] Define the initial blade angle as θ i (t) and the initial water flow channel opening size is S i (t) as a regulatory action;

[0171] The target energy output value is determined based on the power station design capacity, the real-time demand of the power grid dispatching center, the environmental conditions of the current water level data and water flow velocity data, and the economic costs and benefits;

[0172] According to the difference between the actual energy output and the target energy output, the initial reward is calculated as follows:

[0173]

[0174] Among them, r init is the initial reward value, P tg is the target energy output value;

[0175] When init When >0, the reward is positive, indicating that the current energy output is higher than the target energy output, the current state is ideal, and a positive reward is given;

[0176] When init When <0, the reward is negative, indicating that the current energy output is lower than the target energy output, the current state needs to be improved, and a negative reward is given;

[0177] When init When =0, the reward is zero, indicating that the current energy output has just reached the target, the current state has reached the expected target, and no reward or penalty is given.

[0178] It should be noted that this step optimizes the actual energy output of the hydropower station through a reinforcement learning model, calculates the current energy output based on sensor data and compares it with the target value, adjusts the blade angle and water flow channel opening to maximize the reward, and ensures that the power station operates efficiently and economically.

[0179] S4.1. Examples:

[0180] Assume there is a small hydroelectric power station, whose main parameters include: current water level, water flow velocity, blade angle, water channel opening size, actual energy output, target energy output;

[0181] The current water level is defined as the reservoir water level H(t) monitored in real time;

[0182] Water flow velocity is the water flow velocity V(t) monitored in real time;

[0183] The blade angle is the angle θ(t) of the turbine blades;

[0184] The water flow channel opening size is the water flow channel opening size S(t);

[0185] The actual energy output P(t) is calculated;

[0186] Based on physical conditions and actual conditions, the empirical coefficient α of the current water level is set to 0.1kW / m and the empirical coefficient β of the water flow velocity is set to 20kW / (m / s);

[0187] The target energy output is determined based on the power station design capacity, the real-time demand of the power grid dispatch center, the environmental conditions of the current water level data and water flow velocity data, and the economic costs and benefits;

[0188] Define the state as s t , which includes the current water level and water flow speed;

[0189] Define the action as a t , which includes adjusting the blade angle and adjusting the size of the water flow channel opening;

[0190] Through the analysis of the data, the target energy output P is determined. tg =500kW;

[0191] Assume that at a certain time point t, the data measured by the sensor is:

[0192] Current water level = 10m;

[0193] Water flow speed = 2m / s;

[0194] Substitute the formula to calculate the actual energy output:

[0195] P(t)=0.1×10+20×2=41kW;

[0196] Calculate the initial reward:

[0197]

[0198] In order to make the initial reward value more reasonable, the normalization method is used to ensure that the initial reward value is between positive 1 and negative 1. The expression is:

[0199]

[0200] Substituting the obtained parameters,

[0201] The final reward value is -1, which means that the current state needs to be improved.

[0202] S3.2, energy optimization through reinforcement learning model means,

[0203] Select the Deep Q-Network reinforcement learning model;

[0204] Collect historical data and real-time data to build training data sets;

[0205] Initialize Deep Q-Network model parameters;

[0206] The Deep Q-Network model selects and adjusts actions based on the current state;

[0207] Perform adjustment actions and observe the state and actual energy output at the next time;

[0208] According to the next actual energy output and target energy output, the optimized reward is calculated as follows:

[0209]

[0210] Where P(t+1) is the energy output value actually measured at time t+1;

[0211] When t When >0, it means that the actual energy output at the next time is higher than the target energy output, indicating that the adjustment action is ideal and a positive reward is given;

[0212] When t When <0, it means that the actual energy output at the next time is lower than the target energy output, indicating that the execution of the adjustment action needs to be improved and a penalty is given;

[0213] When t=0, it means that the actual energy output at the next time is exactly equal to the target energy output, indicating that the adjustment action has achieved the expected goal, and no reward or penalty is given;

[0214] Store the current state, action, reward, and next state in the experience replay buffer in the Deep Q-Network model;

[0215] Randomly sample a batch of experience data from the experience replay buffer and update the Q value of the Deep Q-Network reinforcement learning model;

[0216] The Deep Q-Network model achieves optimization by continuously learning to make the actual energy output equal to the target energy output.

[0217] S5. Execute the control strategy according to the optimization results, adjust the blade angle and the opening size of the water flow channel, and achieve the optimal energy output of the hydropower station.

[0218] According to the optimized energy, the angle θ(t) of the turbine blade is adjusted, and the expression is:

[0219] θ(t)=γP(t)+δ;

[0220] Among them, γ is the influence coefficient of energy output on blade angle, and δ is the setting value of blade angle;

[0221] According to the water flow velocity and blade angle, adjust the water flow channel opening size S(t), the expression is:

[0222] S(t)=ηV(t)+ζθ(t);

[0223] Among them, η is the empirical coefficient of the influence of water flow velocity on the opening size of the water flow channel, and ζ is the empirical coefficient of the influence of blade angle on the opening size of the water flow channel.

[0224] It should be noted that this step implements a control strategy by adjusting the turbine blade angle and the size of the water channel opening according to the optimized energy output to achieve efficient operation of the hydropower station and maximize energy output.

[0225] S5.1. Examples:

[0226] Assume γ=0.01, δ=5, η=0.02, ζ=0.05;

[0227] Adjust the blade angle and substitute the value into the formula to calculate the angle value:

[0228] θ(t)=0.01×41+5=5.41;

[0229] Adjust the opening size of the water flow channel and substitute the value into the formula to calculate the opening size of the water flow channel:

[0230] S(t)=0.02×2+0.05×5.41=0.3105m;

[0231] The blade angles and water channel opening sizes were adjusted based on calculations, enabling efficient operation of the hydropower station and maximizing energy output.

[0232] This embodiment also provides a monitoring and control system for a hydropower station, including: a calibration module, a processing module, a data analysis module, an optimization module and a control execution module; the calibration module is used to calibrate a variety of sensors and collect data, upload the data of a variety of sensors to the cloud in real time through the Internet of Things LoRaWAN network, and set a data receiving and storage unit in the cloud; the processing module is used to pre-process the raw data received from different sensors, and fuse the pre-processed data using a simple averaging method;

[0233] The data analysis module is used to analyze the historical data and real-time data of the fused sensor data using the Hadoop big data processing framework to generate a comprehensive evaluation report on the operating status of the hydropower station; the optimization module is used to calculate the actual energy output of the hydropower station based on the comprehensive evaluation report and perform energy optimization through the reinforcement learning model; the control execution module is used to execute the control strategy according to the optimization results to achieve the optimal energy output of the hydropower station.

[0234] This embodiment also provides a computer device, which is suitable for the monitoring and control method of a hydropower station, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the monitoring and control method of a hydropower station proposed in the above embodiment.

[0235] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0236] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for monitoring and controlling a hydropower station as proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0237] In summary, the present invention uses the DQN reinforcement learning model to calculate the actual energy output based on real-time sensor data and compare it with the target energy output to generate a reward value. The DQN model selects the optimal action based on the current state and gradually optimizes the control strategy by continuously updating the Q value. This method not only realizes adaptive control, but also can dynamically adjust in complex and changing environments to achieve optimal energy output.

[0238] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for monitoring and controlling a hydropower station, characterized in that: include, Calibrate and collect data from various sensors, upload the data from various sensors to the cloud in real time through the IoT LoRaWAN network, and set up data receiving and storage units in the cloud; Preprocess the raw data received from different sensors and fuse the preprocessed data using a simple averaging method; The fused sensor data is analyzed using the Hadoop big data processing framework to analyze historical and real-time data and generate a comprehensive assessment report on the operating status of the hydropower station; Based on the comprehensive assessment report, the actual energy output of the hydropower station is calculated and energy optimization is performed through the reinforcement learning model; The control strategy is executed according to the optimization results to achieve the optimal energy output of the hydropower station.

2. The method for monitoring and controlling a hydropower station according to claim 1, characterized in that: The method of calibrating and collecting data from multiple sensors, uploading the data from multiple sensors to the cloud in real time through the LoRaWAN network of the Internet of Things, and setting up a data receiving and storage unit in the cloud, specifically includes the following steps: Calibrate water machine flow meters, water level sensors, pressure sensors and temperature sensors; Upload calibrated sensor data to the cloud in real time via the LoRaWAN network; Use AWS IoT Core to create an IoT core in the cloud; Choose Amazon DynamoDB database as the repository for structured data; Create a DynamoDB table in the AWS console and set the read and write capacity units; Define the primary key device string type, timestamp numeric type, data numeric type, status string type, and location string type in the DynamoDB table; Convert sensor data into a unified JSON format and store it in an Amazon DynamoDB database; Use AWS SDK to store the received data in a DynamoDB table; Create a global secondary GSI index for frequently accessed fields; Generate unique certificates and keys for sensor devices; Enable the static encryption feature of DynamoDB to encrypt the stored data using the KMS key; Use IAM policies to strictly control access to DynamoDB tables; Configure the AWS IoT rule engine to send SNS notifications when water level, flow, temperature, and device status change.

3. The method for monitoring and controlling a hydropower station according to claim 2, characterized in that: The raw data received from different sensors are preprocessed, and the data from multiple sensors are fused using a simple averaging method. The specific steps are: The IQR interquartile range method is used to detect and remove data outliers from the water machine flow meter data, the spline interpolation method is used to fill the missing values ​​of the data, and the Savitzky-Golay filter is used to smooth the data to remove the noise of the water machine flow meter data; The Z-Score statistical method is used to detect and remove data outliers from the water level sensor data, the linear interpolation method is used to fill the missing values ​​of the data, and the five-point moving average method is used to smooth the data to remove the noise of the water level sensor data; The local outlier factor (LOF) algorithm is used to detect and remove data outliers from the pressure sensor data. The missing values ​​of the data are filled by the Lagrange interpolation method. The median filter is used to smooth the data and remove the noise of the pressure sensor data. The DBSCAN clustering algorithm is used to detect and remove data outliers from the temperature sensor data, the polynomial interpolation method is used to fill the missing values ​​of the data, and the EWMA exponentially weighted moving average method is used to smooth the data and remove the noise of the temperature sensor data; All preprocessed sensor data are fused through a simple averaging method to obtain fused sensor data.

4. The method for monitoring and controlling a hydroelectric power station according to claim 3, characterized in that: The fused sensor data is analyzed using the Hadoop big data processing framework to analyze historical data and real-time data to generate a comprehensive evaluation report on the operating status of the hydropower station. The specific steps are: The collected real-time data is processed in real time using Apache Kafka and Apache Flink, tools in the Hadoop big data processing framework, to generate an evaluation report on the power plant operation status of real-time data, and to display the real-time data and analysis results through charts and dashboards; Extract historical data from the cloud database, use Hadoop MapReduce to perform batch analysis on the historical data, and use the ARIMA model combined with the LSTM neural network to predict future water level trends; Combine real-time data with historical data analysis results to generate a comprehensive assessment report on the power plant's operating status.

5. The method for monitoring and controlling a hydropower station according to claim 4, characterized in that: Based on the comprehensive evaluation report, the actual energy output of the hydropower station is calculated, and energy optimization is performed through the reinforcement learning model. The specific steps are: Extract the current water level data H(t) and water flow velocity data V(t) from the comprehensive evaluation report, and calculate the actual energy output P(t) at the current time. The expression is: P(t)=αH(t)+βV(t); Among them, t is the time point, α is the empirical coefficient of the current water level, and β is the empirical coefficient of the water flow velocity; Define the current water level data and water flow speed data as the current state; Based on the actual energy output calculated from the current water level data and water flow velocity, the initial blade angle and water flow channel opening size are generated; Define the initial blade angle as θ i (t) and the initial water channel opening size is S i (t) as a regulatory action; The target energy output value is determined based on the power station design capacity, the real-time demand of the power grid dispatching center, the environmental conditions of the current water level data and water flow velocity data, and the economic costs and benefits; According to the difference between the actual energy output and the target energy output, the initial reward is calculated as follows: Among them, r init is the initial reward value, P tg is the target energy output value.

6. The method for monitoring and controlling a hydropower station according to claim 5, characterized in that: The energy optimization through the reinforcement learning model means that Select the Deep Q-Network reinforcement learning model; Collect historical data and real-time data to build training data sets; Initialize Deep Q-Network model parameters; The Deep Q-Network model selects and adjusts actions based on the current state; Perform adjustment actions and observe the state and actual energy output at the next time; According to the next actual energy output and target energy output, the optimized reward is calculated as follows: Where P(t+1) is the energy output value actually measured at time t+1; Store the current state, action, reward, and next state in the experience replay buffer in the Deep Q-Network model; Randomly sample a batch of experience data from the experience replay buffer and update the Q value of the Deep Q-Network reinforcement learning model; The Deep Q-Network model achieves optimization by continuously learning to make the actual energy output equal to the target energy output.

7. The method for monitoring and controlling a hydroelectric power station according to claim 6, characterized in that: The control strategy is executed according to the optimization result to achieve the optimal energy output of the hydropower station. The specific steps are: According to the optimized energy, the angle θ(t) of the turbine blade is adjusted, and the expression is: θ(t)=γP(t)+δ; Among them, γ is the influence coefficient of energy output on blade angle, and δ is the setting value of blade angle; According to the water flow velocity and blade angle, adjust the water flow channel opening size S(t), the expression is: S(t)=ηV(t)+ζθ(t); Among them, η is the empirical coefficient of the influence of water flow velocity on the opening size of the water flow channel, and ζ is the empirical coefficient of the influence of blade angle on the opening size of the water flow channel.

8. A monitoring and control system for a hydropower station, based on the monitoring and control method for a hydropower station according to any one of claims 1 to 7, characterized in that: Including, calibration module, processing module, data analysis module, optimization module and control execution module; The calibration module is used to calibrate various sensors and collect data, upload various sensor data to the cloud in real time through the IoT LoRaWAN network, and set up a data receiving and storage unit in the cloud; The processing module is used to pre-process the raw data received from different sensors and fuse the pre-processed data using a simple averaging method; The data analysis module is used to analyze the historical data and real-time data using the Hadoop big data processing framework after the fusion of sensor data, and generate a comprehensive evaluation report on the operation status of the hydropower station; The optimization module is used to calculate the actual energy output of the hydropower station based on the comprehensive evaluation report and perform energy optimization through a reinforcement learning model; The control execution module is used to execute the control strategy according to the optimization result to achieve the optimal energy output of the hydropower station.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the hydropower station monitoring and control method described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the hydropower station monitoring and control method according to any one of claims 1 to 7 are implemented.

Images