LSTM (Long Short Term Memory) intelligent energy-saving optimization system for collecting ring of steam turbine generator

Abstract

The invention discloses a steam turbine generator collecting ring LSTM intelligent energy-saving optimization system, and belongs to the technical field of intelligent control of power generation equipment. In order to solve the technical problems of overhigh energy consumption and unstable operation caused by insufficient time sequence characteristic capture and dynamic adjustment lag of a steam turbine generator collecting ring in a traditional control method, the invention provides a collecting ring multi-source data sensing system. Multivariable time series data including temperature gradient, contact resistance, carbon brush abrasion rate and environment temperature and humidity are collected in real time, a memory unit state updating mechanism of the LSTM network is utilized, a mapping relation between operation parameters and energy consumption indexes is dynamically established, and therefore the optimal operation parameter combination is predicted. And the system dynamically adjusts the carbon brush contact pressure, the cooling air volume and the voltage parameter based on the prediction result through the optimization strategy generation module, and executes an adjustment instruction through the closed-loop control module. The method is suitable for intelligent operation and maintenance of thermal power, nuclear power and other scenes.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for generators in power systems, specifically to an intelligent energy-saving optimization system for a steam turbine generator slip ring LSTM. Background Technology

[0002] As a core component of energy transmission, the slip ring of a steam turbine generator directly affects the overall energy consumption and stability of the generator set. Traditional slip ring operation strategies are mainly based on fixed control parameters or empirical threshold adjustments, which are difficult to adapt to dynamic load changes and complex operating conditions. Existing technologies have the following limitations: First, conventional control methods such as PID regulation or manual empirical rules cannot capture the nonlinear changes in slip ring contact resistance in real time, leading to increased carbon brush wear and higher energy losses. Second, existing energy consumption optimization schemes mostly rely on offline data analysis and lack the ability to adaptively model multi-dimensional time-series characteristics such as temperature gradients and vibration spectra, making it impossible to achieve dynamic predictive adjustments of operating parameters. Third, maintenance strategies typically adopt a periodic inspection mode, failing to establish a correlation model between equipment degradation status and energy consumption, posing a risk of over-maintenance or delayed fault warnings. Fourth, the application of long short-term memory networks in other industrial scenarios is concentrated on predictive maintenance or fault diagnosis, and has not yet been deeply integrated with the energy control of the slip ring system; the existing model architecture cannot effectively handle the strongly coupled time-series characteristics of the slip ring system.

[0003] Furthermore, existing energy-saving optimization methods often exhibit control delays and overshoot when dealing with sudden changes in grid load or environmental parameter disturbances, leading to abnormal increases in localized collector ring temperature or excessive fluctuations in contact pressure. These problems severely restrict the improvement of energy efficiency and operational economy of steam turbine generator sets, necessitating the development of intelligent optimization methods with self-learning capabilities. Summary of the Invention

[0004] To address the technical problems of excessive energy consumption and unstable operation caused by insufficient timing feature capture and dynamic adjustment lag in traditional control methods for steam turbine generator slip rings, this invention provides a steam turbine generator slip ring LSTM intelligent energy-saving optimization system, comprising:

[0005] The data acquisition module is used to collect the operating data of the slip ring;

[0006] The data processing module is used to preprocess the running data and construct a time series dataset;

[0007] The LSTM core module is used to predict energy consumption on the time series dataset and obtain prediction results.

[0008] An optimization strategy generation module is used to generate optimization control instructions based on the prediction results;

[0009] The closed-loop control module is used to execute the optimized control commands;

[0010] An energy-saving effect evaluation module is used to monitor the actual energy consumption of the closed-loop control module and evaluate the energy-saving effect.

[0011] The system expansion interface is bidirectionally connected to the data processing module, LSTM core module, and energy-saving effect evaluation module, and is used to provide external sensor expansion, algorithm update, and data output functions.

[0012] Furthermore, the data acquisition module includes:

[0013] A collector ring surface temperature sensor array is used to measure the surface temperature of the collector ring.

[0014] A carbon brush contact current sensor is used to measure the current value at the brush contact point.

[0015] Inter-ring voltage difference sensor, used to measure the voltage difference between collector rings;

[0016] Rotor speed sensor, used to measure rotor speed;

[0017] An ambient temperature and humidity sensor is used to measure the temperature and humidity of the environment surrounding the collector ring.

[0018] Furthermore, the data processing module includes:

[0019] The signal conditioning unit is used to filter and amplify the sensor signal;

[0020] The data normalization unit is used to normalize data of different dimensions to the same scale. The normalization formula is:

[0021]

[0022] In the formula The characteristic mean, Standard deviation;

[0023] The time series building unit is used to construct a time series dataset that can be used by the model by arranging normalized data in chronological order.

[0024] Furthermore, the LSTM core module includes an input gate, a forget gate, an output gate, and a memory unit, and its state update mechanism is defined by the following set of formulas:

[0025]

[0026]

[0027]

[0028]

[0029]

[0030] in, This represents the input gate vector at time t. This represents the forgetting gate vector at time t. This represents the output gate vector at time t. This represents the state vector of the memory cell at time t. Let represent the hidden state vector at time t. Let represent the input feature vector at time t. For the sigmoid function, This indicates element-wise multiplication. This is the weight matrix. This is a bias term.

[0031] Furthermore, the optimization strategy generation module includes:

[0032] Fully connected layers are used to map the LSTM hidden states to a 3D control space, as shown in the formula:

[0033]

[0034] in, Indicates the output vector. Indicates the carbon brush pressure setting value, This indicates the cooling airflow setting value. Indicates the dynamic load factor;

[0035] The constraint element is used to set the boundary constraints for control parameters, and its objective function is:

[0036]

[0037] in These are the weighting coefficients. To predict energy consumption, This is a voltage regulation command. For cooling control parameters, the constraints are satisfied as follows:

[0038]

[0039]

[0040]

[0041] in, This is the current measured value of the collector ring voltage. This is the absolute minimum safe operating voltage allowed for the slip ring system. This is the highest safe operating voltage allowed for the slip ring system. This refers to the real-time operating temperature of the slip ring. This is the highest safe temperature allowed by the collector ring material structure.

[0042] Furthermore, the closed-loop control module includes:

[0043] A carbon brush pressure regulating mechanism is used to dynamically adjust the contact pressure between the carbon brush and the slip ring via a hydraulic servo system.

[0044] Cooling fan speed controller, used to adjust cooling air volume using frequency conversion control strategy;

[0045] The PLC controller is used to receive the optimization control command issued by the optimization strategy generation module, convert it into a 4-20 mA analog control signal, and output it to the carbon brush pressure regulating mechanism and the cooling fan speed controller respectively, so as to realize the precise execution of the command and the system linkage control.

[0046] Furthermore, the energy-saving effect evaluation module includes:

[0047] The energy consumption calculation unit is used to collect the actual energy consumption data after the closed-loop control module is executed in real time. The specific formula is as follows:

[0048]

[0049] in This is actual energy consumption data. The time period for integration. Let be the contact loss power at time t. Let t be the power consumption of the cooling system.

[0050] The efficiency improvement assessment unit is used to calculate the relative energy saving rate based on the actual energy consumption data, and the specific formula is as follows:

[0051]

[0052] The baseline value represents the relative energy saving rate. Take energy consumption data under traditional control strategies, To optimize energy consumption.

[0053] Furthermore, the system also includes a system expansion interface, which includes:

[0054] External sensor access port, used to expand the connection of additional sensors;

[0055] An algorithm update interface is used to receive external instructions and update the algorithm of the LSTM core module online.

[0056] The data visualization interface is used to output internal system data to external display devices.

[0057] The beneficial effects of this invention are:

[0058] (1) Establish a spatiotemporal characteristic mapping model of the operating state of the collector ring and capture the accumulation process of current thermal effect through LSTM network;

[0059] (2) It can reduce the energy consumption of the collector ring system and extend the service life of key components;

[0060] (3) Compared with the traditional PID control method, the strategy response speed of this scheme is improved under transient conditions and the accuracy of abnormal condition identification is higher. Attached Figure Description

[0061] Figure 1 This is a diagram showing the module architecture of the present invention. Detailed Implementation

[0062] The technical solution of the present invention will be further described below with reference to embodiments, but it is not limited thereto. Any modifications or equivalent substitutions to the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention should be covered within the protection scope of the present invention. In the following embodiments, the process equipment or apparatus not specifically specified are all conventional equipment or apparatus in the art. Unless otherwise specified, the raw materials used in the embodiments of the present invention are all commercially available; unless otherwise specified, the technical means used in the embodiments of the present invention are all conventional means well known to those skilled in the art.

[0063] Example 1, combined with Figure 1 This embodiment describes, 1. a steam turbine generator slip ring LSTM intelligent energy-saving optimization system, characterized in that it includes:

[0064] The data acquisition module 100 is used to collect the operating data of the slip ring;

[0065] Data processing module 200 is used to preprocess the running data and construct a time series dataset;

[0066] The LSTM core module 300 is used to predict energy consumption on the time series dataset and obtain prediction results.

[0067] The optimization strategy generation module 400 is used to generate optimization control instructions based on the prediction results;

[0068] Closed-loop control module 500 is used to execute the optimized control command;

[0069] An energy-saving effect evaluation module 600 is used to monitor the actual energy consumption of the closed-loop control module 500 and evaluate the energy-saving effect.

[0070] The system expansion interface 700 is bidirectionally connected to the data processing module 200, the LSTM core module 300, and the energy-saving effect evaluation module 600, and is used to provide external sensor expansion, algorithm update, and data output functions.

[0071] Specifically, this invention constructs a data-driven intelligent closed-loop control system. The system uses a data acquisition module to perceive the real-time operating status of the slip ring. After data processing, the data is normalized, and the LSTM core module accurately predicts energy consumption trends. Then, the optimization strategy generation module dynamically generates optimal control commands based on the prediction results, which are executed by the closed-loop control module to adjust parameters such as carbon brush pressure and cooling airflow, ultimately achieving precise energy savings. Simultaneously, the energy-saving effect evaluation module continuously monitors actual energy consumption and feeds the evaluation results back to the front-end strategy module, forming an adaptive loop that can continuously optimize itself based on actual results. This effectively improves the slip ring's operating efficiency and extends equipment lifespan while ensuring equipment safety.

[0072] The data acquisition module 100 includes:

[0073] The slip ring surface temperature sensor array 101 is used to measure the surface temperature of the slip ring.

[0074] The carbon brush contact current sensor 102 is used to measure the current value at the brush contact point.

[0075] Inter-ring voltage difference sensor 103 is used to measure the voltage difference between collector rings;

[0076] Rotor speed sensor 104 is used to measure rotor speed;

[0077] The ambient temperature and humidity sensor 105 is used to measure the temperature and humidity of the environment around the collector ring.

[0078] The data processing module 200 includes:

[0079] The signal conditioning unit 201 is used to filter and amplify the sensor signal;

[0080] Data normalization unit 202 is used to normalize data of different dimensions to the same scale. The normalization formula is as follows:

[0081]

[0082] In the formula The characteristic mean, Standard deviation;

[0083] The time series construction unit 203 is used to construct a time series dataset that can be used by the model by arranging the normalized data in chronological order.

[0084] The data processing module uses the Z-score normalization method to eliminate the dimensional differences and extreme fluctuations of each feature value, scaling them to a similar numerical scale to ensure the stability and convergence speed of the LSTM model training. Next, the time series construction unit reorganizes the normalized static data points into a continuous time series sample set in chronological order within a sliding window. The purpose is to mine and preserve the dynamic change patterns and causal relationships contained in the data, providing a structured data foundation for the LSTM core module to perform time series feature learning and energy consumption prediction.

[0085] The LSTM core module 300 includes an input gate 301, a forget gate 302, an output gate 303, and a memory unit 304. Its state update mechanism is defined by the following set of formulas:

[0086]

[0087]

[0088]

[0089]

[0090]

[0091] in, This represents the input gate vector at time t. This represents the forgetting gate vector at time t. This represents the output gate vector at time t. This represents the state vector of the memory cell at time t. Let represent the hidden state vector at time t. Let represent the input feature vector at time t. For the sigmoid function, This indicates element-wise multiplication. This is the weight matrix. This is a bias term.

[0092] The core principle of LSTM lies in simulating and managing the dynamic learning and memory of long-term dependencies in time-series data through the synergistic effect of input gates, forget gates, output gates, and memory units. Its fundamental purpose is to solve the gradient vanishing or exploding problem in traditional recurrent neural networks (RNNs), thereby accurately capturing the temporal dynamic characteristics and nonlinear changes in collector loop operating parameters (such as temperature and current).

[0093] The optimization strategy generation module 400 includes:

[0094] Fully connected layer 401 is used to map the LSTM hidden states to the 3D control space, as shown in the formula:

[0095]

[0096] in, Indicates the output vector. Indicates the carbon brush pressure setting value, This indicates the cooling airflow setting value. Indicates the dynamic load factor;

[0097] Constraint element 402 is used to set boundary constraints for control parameters, and its objective function is:

[0098]

[0099] in These are the weighting coefficients. To predict energy consumption, This is a voltage regulation command. For cooling control parameters, the constraints are satisfied as follows:

[0100]

[0101]

[0102]

[0103] in, This is the current measured value of the collector ring voltage. This is the absolute minimum safe operating voltage allowed for the slip ring system. This is the highest safe operating voltage allowed for the slip ring system. This refers to the real-time operating temperature of the slip ring. This is the highest safe temperature allowed by the collector ring material structure.

[0104] The principle of the optimization strategy generation module lies in transforming the prediction information from the LSTM core module into executable control instructions. Its core is a constrained multi-objective optimization process. This module uses a fully connected layer to initially decode the hidden features representing the future state of the system into an initial vector of control parameters. Subsequently, the constraint unit, with the core objective of minimizing predicted energy consumption while ensuring the smoothness of control actions, modifies and optimizes the initial instructions under the strict conditions of voltage safety range and absolute temperature safety limits, ultimately generating the optimal control instructions that pursue energy-saving benefits while absolutely guaranteeing equipment safety.

[0105] The closed-loop control module 500 includes:

[0106] The carbon brush pressure regulating mechanism 501 is used to dynamically adjust the contact pressure between the carbon brush and the slip ring through a hydraulic servo system.

[0107] Cooling fan speed controller 502 is used to adjust the cooling air volume using a frequency conversion control strategy;

[0108] The PLC controller 503 is used to receive the optimization control command issued by the optimization strategy generation module 400, convert it into a 4-20 mA analog control signal, and output it to the carbon brush pressure regulating mechanism 501 and the cooling fan speed controller 502 respectively, so as to realize the precise execution of the command and the system linkage control.

[0109] The energy-saving effect evaluation module 600 includes:

[0110] Energy consumption calculation unit 601 is used to collect the actual energy consumption data after the closed-loop control module 500 executes in real time. The specific formula is as follows:

[0111]

[0112] in This is actual energy consumption data. The time period for integration. Let be the contact loss power at time t. Let t be the power consumption of the cooling system.

[0113] The efficiency improvement assessment unit 602 is used to calculate the relative energy saving rate based on the actual energy consumption data, and the specific formula is as follows:

[0114]

[0115] The baseline value represents the relative energy saving rate. Take energy consumption data under traditional control strategies, To optimize energy consumption.

[0116] The energy-saving effect evaluation module accurately measures the actual energy consumption after the closed-loop control module is executed by the energy consumption calculation unit. The efficiency improvement evaluation unit compares the optimized energy consumption with the benchmark energy consumption to calculate the objective relative energy saving rate. This quantitatively verifies the actual energy-saving effect of the intelligent optimization strategy and provides key decision-making basis for whether the system needs to trigger model updates or strategy adjustments, ultimately ensuring that the entire optimization system can operate continuously and effectively.

[0117] The system also includes a system expansion interface 700, which includes:

[0118] External sensor access port 701 is used to expand the connection of additional sensors;

[0119] The algorithm update interface 702 is used to receive external instructions and update the algorithm of the LSTM core module 300 online.

[0120] The data visualization interface 703 is used to output internal system data to an external display device.

Claims

1. A steam turbine generator slip ring LSTM intelligent energy-saving optimization system, characterized in that, include: The data acquisition module (100) is used to collect the operating data of the collector ring; The data processing module (200) is used to preprocess the running data and construct a time series dataset; The LSTM core module (300) is used to predict energy consumption on the time series dataset and obtain prediction results; An optimization strategy generation module (400) is used to generate optimization control instructions based on the prediction results; The closed-loop control module (500) is used to execute the optimized control command; An energy-saving effect evaluation module (600) is used to monitor the actual energy consumption of the closed-loop control module (500) and evaluate the energy-saving effect; The system expansion interface (700) is bidirectionally connected to the data processing module (200), the LSTM core module (300), and the energy-saving effect evaluation module (600) to provide external sensor expansion, algorithm update, and data output functions.

2. The turbine generator slip ring LSTM intelligent energy-saving optimization system according to claim 1, characterized in that, The data acquisition module (100) includes: A collector ring surface temperature sensor array (101) is used to measure the surface temperature of the collector ring; A carbon brush contact current sensor (102) is used to measure the current value at the brush contact point; Inter-ring voltage difference sensor (103) is used to measure the voltage difference between collector rings; Rotor speed sensor (104) is used to measure rotor speed; An ambient temperature and humidity sensor (105) is used to measure the temperature and humidity of the environment around the collector ring.

3. The turbine generator slip ring LSTM intelligent energy-saving optimization system according to claim 1, characterized in that, The data processing module (200) includes: The signal conditioning unit (201) is used to filter and amplify the sensor signal; The data normalization unit (202) is used to normalize data of different dimensions to the same scale. The normalization formula is as follows: In the formula The characteristic mean, Standard deviation; The time series building unit (203) is used to build the normalized data into a time series dataset that can be used by the model in chronological order.

4. The turbine generator slip ring LSTM intelligent energy-saving optimization system according to claim 1, characterized in that, The LSTM core module (300) includes an input gate (301), a forget gate (302), an output gate (303), and a memory unit (304), and its state update mechanism is defined by the following set of formulas: in, This represents the input gate vector at time t. This represents the forgetting gate vector at time t. This represents the output gate vector at time t. This represents the state vector of the memory cell at time t. Let represent the hidden state vector at time t. Let represent the input feature vector at time t. For the sigmoid function, This indicates element-wise multiplication. This is the weight matrix. This is a bias term.

5. The turbine generator slip ring LSTM intelligent energy-saving optimization system according to claim 1, characterized in that, The optimization strategy generation module (400) includes: The fully connected layer (401) is used to map the LSTM hidden states to the 3D control space, as shown in the formula: in, Indicates the output vector. Indicates the carbon brush pressure setting value, This indicates the cooling airflow setting value. Indicates the dynamic load factor; The constraint element (402) is used to set the boundary constraints of the control parameters, and its objective function is: in These are the weighting coefficients. To predict energy consumption, This is a voltage regulation command. For cooling control parameters, the constraints are satisfied as follows: in, This is the current measured value of the collector ring voltage. This is the absolute minimum safe operating voltage allowed for the slip ring system. This is the highest safe operating voltage allowed for the slip ring system. This refers to the real-time operating temperature of the slip ring. This is the highest safe temperature allowed by the collector ring material structure.

6. The turbine generator slip ring LSTM intelligent energy-saving optimization system according to claim 1, characterized in that, The closed-loop control module (500) includes: A carbon brush pressure regulating mechanism (501) is used to dynamically adjust the contact pressure between the carbon brush and the slip ring through a hydraulic servo system. Cooling fan speed controller (502) is used to adjust the cooling air volume using a frequency conversion control strategy; The PLC controller (503) is used to receive the optimization control command issued by the optimization strategy generation module (400), convert it into a 4-20 mA analog control signal, and output it to the carbon brush pressure regulating mechanism (501) and the cooling fan speed controller (502) respectively, so as to realize the precise execution of the command and the system linkage control.

7. The turbine generator slip ring LSTM intelligent energy-saving optimization system according to claim 1, characterized in that, The energy-saving effect evaluation module (600) includes: The energy consumption calculation unit (601) is used to collect the actual energy consumption data after the closed-loop control module (500) is executed in real time. The specific formula is as follows: in This is actual energy consumption data. The time period for integration. Let be the contact loss power at time t. The power consumption of the cooling system at time t; The efficiency improvement assessment unit (602) is used to calculate the relative energy saving rate based on the actual energy consumption data, and the specific formula is as follows: The baseline value represents the relative energy saving rate. Take energy consumption data under traditional control strategies, To optimize energy consumption.

8. The turbine generator slip ring LSTM intelligent energy-saving optimization system according to claim 1, characterized in that, The system also includes a system expansion interface (700), which includes: External sensor access port (701) is used to expand the access of additional sensors; The algorithm update interface (702) is used to receive external instructions and update the algorithm of the LSTM core module (300) online; The data visualization interface (703) is used to output internal system data to an external display device.

Images