Steam extraction energy storage and heat energy storage integrated regulation control method and system

Abstract

The invention provides a steam extraction energy storage and heat energy storage integrated regulation control method and system, and relates to the technical field, and the method comprises the steps: building a digital twinborn model of a steam extraction energy storage and heat energy storage integrated system, and respectively obtaining corresponding data; establishing a multi-objective optimization model based on the collected data; solving the multi-objective optimization model by using an improved multi-objective differential evolution algorithm in combination with an adaptive variation strategy and a non-dominated sorting mechanism to obtain an optimal adjustment strategy; on the basis of the optimal adjusting strategy, the opening degree of a steam extraction valve of each stage of steam extraction point is adjusted in real time through a distributed control system, and the steam extraction flow is controlled; the inflation power and the deflation power of the compressed air energy storage unit are dynamically adjusted, and meanwhile the charging and discharging power of the flywheel energy storage unit is controlled; the heat charging and discharging rate of the phase change material heat storage unit is adjusted in a self-adaptive mode, and meanwhile the heat charging and discharging flow of the fused salt heat storage unit is controlled.

Description

Technical Field

[0001] The present invention relates to the technical field, and in particular to a method and system for integrated regulation and control of extraction steam energy storage and thermal energy storage. Background Art

[0002] As the proportion of renewable energy in the power system continues to increase, the demand for peak and frequency regulation of the power grid is increasing, and traditional thermal power units are facing challenges such as insufficient flexibility and reduced economy. In order to improve the regulation capability and economy of thermal power units, the integrated system of extraction steam energy storage and thermal energy storage came into being. By combining extraction steam energy storage technology with thermal energy storage technology, electric and thermal decoupling is achieved, which can effectively improve the flexibility and economy of the system.

[0003] At present, the research on integrated systems of extraction steam energy storage and thermal energy storage mainly focuses on system configuration optimization, operation strategy optimization, control system design and economic analysis. However, the existing regulation and control methods still have many shortcomings. The system is highly complex, involving multiple subsystems and the conversion of multiple energy forms, making it difficult to achieve precise control; the dynamic response capability is insufficient, making it difficult to cope with the rapid changes in grid load and thermal load; the prediction capability is insufficient, and the prediction of the future state of the system and changes in the external environment is not accurate enough; the adaptive ability is not strong, making it difficult to automatically adjust the control strategy according to changes in the system operating state and the external environment; the ability to handle abnormal situations is weak, and there are safety hazards; the existing methods are mostly based on a single optimization goal, making it difficult to achieve comprehensive optimization of multiple goals.

[0004] In summary, there is an urgent need for a new integrated regulation and control method for steam extraction energy storage and thermal energy storage, which can effectively handle the high complexity of the system and achieve precise control; has fast dynamic response capabilities; has strong prediction capabilities; has good adaptive capabilities and can automatically adjust the control strategy. At the same time, it improves system safety and the level of intelligent control. The present invention can solve the problems in the prior art. Summary of the invention

[0005] The embodiments of the present invention provide a method and system for integrated regulation and control of extraction steam energy storage and thermal energy storage, which can solve the problems in the prior art.

[0006] According to a first aspect of the embodiments of the present invention,

[0007] Provided is a method for integrated regulation and control of extraction steam energy storage and thermal energy storage, comprising:

[0008] Establish a digital twin model of an integrated system of steam extraction energy storage and thermal energy storage, wherein the digital twin model includes a steam turbine steam extraction system, an energy storage system and a thermal energy storage system; the steam turbine steam extraction system includes a multi-stage steam extraction point model and a steam extraction pipeline model; the energy storage system includes a compressed air energy storage unit model and a flywheel energy storage unit model; the thermal energy storage system includes a phase change material heat storage unit model and a molten salt heat storage unit model; respectively collect a first real-time operating parameter corresponding to the steam turbine steam extraction system, a second real-time operating parameter corresponding to the energy storage system, a third real-time operating parameter corresponding to the thermal energy storage system, as well as a real-time load curve of the power grid, load forecast data for the next 24 hours, ambient temperature change trends and renewable energy power generation forecast information;

[0009] A multi-objective optimization model is established based on the digital twin model of the integrated system of extracted steam storage and thermal energy storage, the first real-time operating parameter, the second real-time operating parameter, the third real-time operating parameter, the real-time load curve of the power grid, the load forecast data for the next 24 hours, the ambient temperature change trend and the renewable energy power generation forecast information; an improved multi-objective differential evolution algorithm is used, combined with an adaptive mutation strategy and a non-dominated sorting mechanism, to solve the multi-objective optimization model and obtain the optimal regulation strategy;

[0010] Based on the optimal regulation strategy, a multi-stage steam extraction flow regulation instruction is sent to the steam turbine extraction system through a distributed control system, and a fuzzy PID control algorithm is used to adjust the opening of the steam extraction valves at each level of extraction points in real time to control the steam extraction flow; a coordinated charging and discharging instruction is sent to the energy storage system, and the charging power and discharging power of the compressed air energy storage unit are dynamically adjusted according to the load changes of the power grid and the electricity price information, and the charging and discharging power of the flywheel energy storage unit is controlled at the same time; a coordinated heat storage instruction is sent to the thermal energy storage system, and the charging and discharging rates of the phase change material heat storage unit are adaptively adjusted according to the heat load demand and the ambient temperature changes, and the charging and discharging flow of the molten salt heat storage unit is controlled at the same time.

[0011] In an optional embodiment,

[0012] A digital twin model of an integrated system of steam extraction energy storage and thermal energy storage is established, wherein the digital twin model includes a steam turbine steam extraction system, an energy storage system and a thermal energy storage system; the steam turbine steam extraction system includes a multi-stage steam extraction point model and a steam extraction pipeline model; the energy storage system includes a compressed air energy storage unit model and a flywheel energy storage unit model; the thermal energy storage system includes a phase change material heat storage unit model and a molten salt heat storage unit model, including:

[0013] A digital twin model of a steam turbine extraction system is established, wherein the digital twin model includes a multi-stage extraction point model and an extraction pipeline model; the multi-stage extraction point model determines the relationship between the pressure and flow rate between the steam turbine stages based on the Stodola ellipse law, and calculates the steam parameters of each stage of extraction points in combination with the heat balance equation; the extraction pipeline model is based on the one-dimensional unsteady flow theory, combined with pipeline friction, heat loss and pressure fluctuation factors, and solves the steam flow state in the pipe through the characteristic line method;

[0014] Establish a digital twin model of an energy storage system, wherein the digital twin model of the energy storage system includes a compressed air energy storage unit model and a flywheel energy storage unit model; the compressed air energy storage unit model is based on the isentropic compression-expansion theory, combined with the actual efficiency curves of the compressor and the turbine, simulates the energy conversion efficiency of the charging and discharging process, and determines the state change of the compressed air through the non-ideal gas state equation; the flywheel energy storage unit model is based on the law of conservation of angular momentum, combined with the dynamics model of the magnetic bearing system, to construct the acceleration process and deceleration process of the flywheel and determine the energy loss;

[0015] A digital twin model of a thermal energy storage system is established, wherein the digital twin model of the thermal energy storage system includes a phase change material heat storage unit model and a molten salt heat storage unit model; the phase change material heat storage unit model adopts the enthalpy-porosity method, combined with the effective thermal conductivity, to simulate the heat transfer and phase change interface movement during the phase change process; the molten salt heat storage unit model is based on the computational fluid dynamics theory, adopts the three-dimensional non-steady-state heat conduction equation, and combines the influence of natural convection and thermal radiation to determine the molten salt temperature field distribution.

[0016] In an optional embodiment,

[0017] The improved multi-objective differential evolution algorithm is used, combined with the adaptive mutation strategy and the non-dominated sorting mechanism, to solve the multi-objective optimization model and obtain the optimal adjustment strategy including:

[0018] Set the population size and maximum number of iterations, randomly generate an initial feasible solution set that meets the constraints as the initial population, use the initial population as the current population, and start iteration:

[0019] Performing non-dominated sorting on the current population, dividing the individuals in the current population into different non-dominated levels, calculating the crowding distance of individuals in the same non-dominated level in the current population, sorting the current population based on the non-dominated level and the crowding distance to obtain an ordered population, calculating the fitness improvement rate of the ordered population, and dynamically adjusting the mutation factor based on the fitness improvement rate;

[0020] Use the adjusted mutation factor to perform mutation operation on each individual in the ordered population, randomly select three different individuals, generate mutant individuals through linear combination to obtain a mutant population, perform binomial crossover between the mutant population and the ordered population to generate a crossover population, and evaluate the objective function value of each individual in the crossover population based on the multi-objective optimization model;

[0021] Merge the ordered population and the crossover population, perform non-dominated sorting and crowding distance calculation on the merged population again, obtain new non-dominated sorting results and new crowding distances, and select the best individuals from the merged population to form a new generation of population based on the new non-dominated sorting results and new crowding distances;

[0022] The new generation population is taken as the current population, and the iteration is repeated until the preset maximum number of iterations is reached. The non-dominated solution set is extracted from the last generation population to determine the Pareto optimal solution set. The fuzzy comprehensive evaluation method is used to combine the decision maker's preferences and actual operating conditions to select the final optimal solution from the Pareto optimal solution set.

[0023] In an optional embodiment,

[0024] Based on the optimal regulation strategy, a multi-stage steam extraction flow rate regulation instruction is sent to the steam turbine extraction system through a distributed control system, and the steam extraction valve opening of each level of steam extraction point is adjusted in real time by a fuzzy PID control algorithm to control the steam extraction flow rate; a coordinated charging and discharging instruction is sent to the energy storage system, and the charging power and discharging power of the compressed air energy storage unit are dynamically adjusted according to the load change of the power grid and the electricity price information, and the charging and discharging power of the flywheel energy storage unit is controlled at the same time; a coordinated heat storage instruction is sent to the thermal energy storage system, and the charging and discharging rate of the phase change material heat storage unit is adaptively adjusted according to the heat load demand and the ambient temperature change, and the charging and discharging flow rate of the molten salt heat storage unit is controlled at the same time, including:

[0025] According to the multi-stage steam extraction flow dynamic regulation curve in the optimal regulation strategy, the actual steam extraction flow of each stage of steam extraction point is collected, the deviation and the deviation change rate from the target steam extraction flow are calculated, and the deviation and the deviation change rate are fuzzy processed to divide them into fuzzy subsets;

[0026] Based on the fuzzy subset, an if-then fuzzy rule base is established, and the centroid method is used for defuzzification to obtain the adjustment amount of the proportional integral differential controller parameters, and the opening adjustment amount of the steam extraction valve is calculated by updating the proportional integral differential controller parameters, and the opening adjustment amount is sent to the corresponding steam extraction valve actuator through the distributed control system;

[0027] Obtaining grid load data and electricity price information in real time, calculating the load deviation and electricity price gradient at the corresponding moment, and dividing the system operation state into multiple operation modes based on the corresponding values ​​of the load deviation and electricity price gradient, including peak-valley filling mode, peak shaving mode, valley filling mode and economic operation mode;

[0028] Based on different operation modes, dynamically adjust the first charging and discharging power of the compressed air energy storage unit and the second charging and discharging power of the flywheel energy storage unit, and send the adjustment instructions of the first charging and discharging power and the second charging and discharging power to the corresponding charging and discharging actuators through the distributed control system;

[0029] Collect heat load demand and ambient temperature in real time, calculate heat load deviation and ambient temperature change rate, and divide the operation state of the thermal energy storage system into a fast response mode, a stable operation mode and an energy-saving mode according to the corresponding values ​​of the heat load deviation and the ambient temperature change rate;

[0030] The first current heat output of the phase change material heat storage unit and the second current heat storage of the molten salt heat storage unit are calculated; according to the ratio of the first current heat storage to the second current heat storage, the charging and discharging power distribution ratio of the phase change material heat storage unit and the molten salt heat storage unit are dynamically adjusted; based on the charging and discharging power distribution ratio, the charging and discharging rate of the phase change material heat storage unit and the charging and discharging flow rate of the molten salt heat storage unit are calculated, and sent to the corresponding charging and discharging actuators through the distributed control system.

[0031] In an optional embodiment,

[0032] The charging and discharging heat power distribution ratio is as follows:

[0033]

[0034] Among them, α represents the charging and discharging heat power allocation ratio, k 1 represents the first weight, S PCM represents the current heat storage capacity of the phase change material heat storage unit, S salt represents the current heat storage capacity of the molten salt heat storage unit, k 2 represents the second weight, S salt-max Indicates the maximum heat storage capacity of the molten salt heat storage unit;

[0035] The heat charging and discharging rate of the phase change material heat storage unit is as follows:

[0036]

[0037] Among them, R PCM represents the charging and discharging rate of the phase change material heat storage unit, ΔQ represents the heat load deviation, and ΔH PCM Represents the latent heat of phase change of phase change material;

[0038] The charging and discharging heat flow of the molten salt heat storage unit is as follows:

[0039]

[0040] Among them, F salt represents the heat charging and discharging flow rate of the molten salt heat storage unit, c p-salt represents the specific heat capacity of the molten salt, ΔT salt Represents the temperature difference of the molten salt.

[0041] In an optional embodiment,

[0042] The method further comprises:

[0043] Setting up multiple sensors, acquiring real-time data from the sensors, performing data cleaning and preprocessing on the real-time data from the sensors to obtain real-time operation data, inputting the real-time operation data into a deep learning model, performing feature extraction on the real-time operation data, outputting the predicted status of each component of the system, and determining the abnormal situation prediction result;

[0044] Based on the abnormal situation prediction results, the cumulative deviation of multiple control parameters is calculated. When the cumulative deviation exceeds the preset control threshold, it is determined that the system operation state has undergone an over-threshold change, triggering the online adaptive optimization mechanism, resetting all control parameters to safe default values, and increasing the sampling frequency of multiple sensors to obtain high-frequency sampling data;

[0045] The digital twin model is updated based on the high-frequency sampling data using the recursive least squares method, and based on the current state of the model and external environment information, the optimization objective function is redefined, the optimal adjustment strategy is re-determined, and the optimal adjustment strategy is simulated and verified. If the simulation results show that the constraints are violated, the optimal adjustment strategy is locally readjusted through the pattern search algorithm to obtain the final optimal adjustment strategy, and the final optimal adjustment strategy is executed through the actuator.

[0046] In an optional embodiment,

[0047] The deep learning model includes:

[0048] Construct an input layer to receive multi-dimensional real-time running data, connect a first convolution layer to the input layer, use a one-dimensional convolution operation in the first convolution layer, determine a small-scale convolution kernel according to a preset small-scale value, set the step size to a minimum value, keep the output and input sizes equal by padding, and apply a ReLU activation function to obtain a first convolution feature map, input the first convolution feature map into a first maximum pooling layer, set the pooling window size and step size of the first maximum pooling layer to the same value, and obtain a first pooling feature map;

[0049] Input the first pooled feature map into the first long short-term memory network layer, extract the time series feature according to the preset number of units to obtain the first long short-term memory feature, input the first long short-term memory feature into the second long short-term memory network layer, the number of units in the second long short-term memory network layer is half of the preset number of units, extract the long-term dependency feature to obtain the second long short-term memory feature, input the second long short-term memory feature into the global average pooling layer, map sequences of different lengths to vectors of preset fixed length, and obtain a global feature vector;

[0050] Input the global feature vector into the first fully connected layer, and obtain a first fully connected feature through a ReLU activation function; input the first fully connected feature into a Dropout layer, and obtain a random dropout feature based on a preset dropout rate; input the random dropout feature into a second fully connected layer, and set the number of neurons in the second fully connected layer to half the number of neurons in the first fully connected layer, and obtain a second fully connected feature through a ReLU activation function;

[0051] The second fully connected feature is input into the output layer, and the probability distribution of each abnormal category is obtained through the Softmax activation function.

[0052] According to a second aspect of the embodiments of the present invention,

[0053] Provided is an integrated regulation and control system for extraction steam energy storage and thermal energy storage, comprising:

[0054] The first unit is used to establish a digital twin model of an integrated system of steam extraction energy storage and thermal energy storage, wherein the digital twin model includes a steam turbine extraction system, an energy storage system and a thermal energy storage system; the steam turbine extraction system includes a multi-stage extraction point model and an extraction pipeline model; the energy storage system includes a compressed air energy storage unit model and a flywheel energy storage unit model; the thermal energy storage system includes a phase change material heat storage unit model and a molten salt heat storage unit model; and respectively collects a first real-time operating parameter corresponding to the steam turbine extraction system, a second real-time operating parameter corresponding to the energy storage system, a third real-time operating parameter corresponding to the thermal energy storage system, and a real-time load curve of the power grid, load forecast data for the next 24 hours, ambient temperature change trends, and renewable energy power generation forecast information;

[0055] The second unit is used to establish a multi-objective optimization model based on the digital twin model of the integrated system of extraction steam storage and thermal energy storage, the first real-time operating parameter, the second real-time operating parameter, the third real-time operating parameter, the real-time load curve of the power grid, the load forecast data for the next 24 hours, the ambient temperature change trend and the renewable energy power generation forecast information; using the improved multi-objective differential evolution algorithm, combined with the adaptive mutation strategy and the non-dominated sorting mechanism, the multi-objective optimization model is solved to obtain the optimal regulation strategy;

[0056] The third unit is used to send multi-stage steam extraction flow adjustment instructions to the steam turbine extraction system through the distributed control system based on the optimal adjustment strategy, and adopt the fuzzy PID control algorithm to adjust the steam extraction valve opening of each level of extraction point in real time to control the steam extraction flow; send coordinated charging and discharging instructions to the energy storage system, dynamically adjust the charging power and discharging power of the compressed air energy storage unit according to the load change of the power grid and the electricity price information, and control the charging and discharging power of the flywheel energy storage unit; send coordinated heat storage instructions to the thermal energy storage system, adaptively adjust the charging and discharging rates of the phase change material heat storage unit according to the heat load demand and the ambient temperature change, and control the charging and discharging flow of the molten salt heat storage unit.

[0057] According to a third aspect of the embodiments of the present invention,

[0058] An electronic device is provided, comprising:

[0059] processor;

[0060] a memory for storing processor-executable instructions;

[0061] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0062] A fourth aspect of the embodiments of the present invention is:

[0063] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0064] In the embodiment of the present invention, accurate monitoring of the operating status is ensured by collecting multi-level real-time parameters of the steam turbine, energy storage and thermal energy storage system; the extraction flow regulation, energy storage charging and discharging and thermal energy storage strategies are optimized by combining information such as grid load forecast and renewable energy power generation forecast; the accuracy and efficiency of the system are improved and energy loss is reduced by dynamically adjusting model parameters; based on future load and power generation forecasts, optimized charging and discharging and thermal energy utilization strategies are formulated in advance to improve the overall energy efficiency of the system; by collecting real-time data and load forecasts, using multi-objective optimization models and digital twin technology, the operating parameters of each subsystem are dynamically adjusted to significantly improve the overall efficiency of the system; the operating points of equipment such as flywheel energy storage and compressed air energy storage are optimized to extend equipment life and reduce electricity purchase and maintenance costs; Energy system and thermal energy storage system store energy during the peak period of renewable energy generation and release it during the off-peak period, thereby increasing the utilization ratio of renewable energy and improving the system's ability to absorb intermittent energy. An improved multi-objective differential evolution algorithm is used to output a variety of optimal adjustment schemes, realize flexible scheduling decisions, and enhance the adaptability and flexibility of the system. The fuzzy PID control algorithm is used to adjust the opening of the steam extraction valve in real time to ensure that the steam extraction flow at each level of the turbine accurately matches the load demand, thereby improving the system's response speed and stability. According to changes in grid load and electricity prices, the charging and discharging strategies of compressed air energy storage and flywheel energy storage units are dynamically optimized to improve the flexibility and economic benefits of the energy storage system. The charging and discharging rates of phase change materials and molten salt heat storage units are adaptively adjusted to ensure that the thermal energy storage system operates efficiently according to demand and maximize the efficiency of thermal energy utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 It is a flow chart of an integrated regulation and control method of extraction steam energy storage and thermal energy storage according to an embodiment of the present invention;

[0066] Figure 2 It is a structural schematic diagram of an integrated regulation and control system of steam extraction energy storage and thermal energy storage according to an embodiment of the present invention. DETAILED DESCRIPTION

[0067] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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.

[0068] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0069] Figure 1 FIG. 1 is a flow chart of an integrated regulation and control method for extraction steam energy storage and thermal energy storage according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0070] S101. Establish a digital twin model of an integrated system of steam extraction energy storage and thermal energy storage, wherein the digital twin model includes a steam turbine steam extraction system, an energy storage system and a thermal energy storage system; the steam turbine steam extraction system includes a multi-stage steam extraction point model and a steam extraction pipeline model; the energy storage system includes a compressed air energy storage unit model and a flywheel energy storage unit model; the thermal energy storage system includes a phase change material heat storage unit model and a molten salt heat storage unit model; respectively collect a first real-time operating parameter corresponding to the steam turbine steam extraction system, a second real-time operating parameter corresponding to the energy storage system, a third real-time operating parameter corresponding to the thermal energy storage system, and a real-time load curve of the power grid, load forecast data for the next 24 hours, ambient temperature change trends, and renewable energy power generation forecast information;

[0071] In order to achieve efficient operation and precise control of the integrated system of steam extraction energy storage and thermal energy storage, a digital twin model of the system is first established. The model includes three main parts: steam turbine extraction system, energy storage system and thermal energy storage system.

[0072] The steam turbine extraction system model consists of a multi-stage extraction point model and an extraction pipeline model. The multi-stage extraction point model describes the pressure, temperature and flow characteristics of each extraction point through thermodynamic equilibrium equations and mass balance equations. The extraction pipeline model considers pipeline pressure loss, heat loss and flow characteristics, and uses a one-dimensional unsteady flow model for simulation. For example, for a 300MW steam turbine, three extraction point models of high pressure, medium pressure and low pressure can be established, and the pressure, temperature and flow range of each extraction point are: high pressure (3.5-4.5MPa, 350-400℃, 50-80t / h), medium pressure (1.5-2.5MPa, 250-300℃, 30-50t / h) and low pressure (0.3-0.5MPa, 150-200℃, 20-40t / h).

[0073] The energy storage system model includes a compressed air energy storage unit model and a flywheel energy storage unit model. The compressed air energy storage unit model takes into account the characteristics of the compressor, gas tank and expander, and uses thermodynamic models and pneumatic models to describe the system's charging and discharging process. The flywheel energy storage unit model is based on rotor dynamics theory to establish detailed models of the motor / generator, bearings and vacuum chamber. Taking a 100MWh compressed air energy storage system as an example, its gas tank volume is 50,000m 3The working pressure range is 4-8MPa, the inflation power is 40MW, and the deflation power is 60MW; the rotor mass of the 10MWh flywheel energy storage system is 20 tons, the maximum speed is 10,000 rpm, and the charging and discharging power is 20MW.

[0074] The thermal energy storage system model consists of a phase change material heat storage unit model and a molten salt heat storage unit model. The phase change material heat storage unit model adopts a heat conduction-phase change coupling model to take into account the nonlinear thermal properties of the material and the latent heat of phase change. The molten salt heat storage unit model is based on computational fluid dynamics theory to establish a detailed model of the molten salt tank, heat exchanger and circulation pump. For example, for a 500MWh thermal energy storage system, a 200MWh phase change material heat storage unit (using paraffin as the phase change material, phase change temperature 60-65°C, phase change latent heat 200kJ / kg) and a 300MWh molten salt heat storage unit (using solar salt, operating temperature range 290-565°C, specific heat capacity 1.5kJ / (kg·K)) can be used.

[0075] After the digital twin model is established, the operating parameters of each subsystem are obtained in real time through the data acquisition system. For the steam turbine extraction system, the pressure, temperature, flow and other parameters of the extraction points at each level are collected, and the sampling frequency is 1Hz. In terms of the energy storage system, the compressed air energy storage unit collects parameters such as the pressure, temperature, compressor power, and expander power of the air storage tank, and the flywheel energy storage unit collects parameters such as speed, charge and discharge power, and bearing temperature, and the sampling frequency is 10Hz. The thermal energy storage system collects parameters such as the temperature distribution, flow, and inlet and outlet temperatures of the phase change material and molten salt, and the sampling frequency is 0.1Hz.

[0076] At the same time, the grid load curve data is obtained in real time from the grid dispatching center, including the current load value and historical data with a resolution of 15 minutes. The load forecast data for the next 24 hours is obtained through the load forecasting system, which comprehensively considers historical load data, weather forecasts, holiday information and other factors, and uses deep learning algorithms for forecasting. The forecast accuracy reaches more than 95% during peak and valley load periods. The environmental temperature change trend data is obtained from the meteorological department, including the current temperature and hourly forecast data for the next 24 hours. The renewable energy power generation forecast information is obtained from the forecasting system of wind farms and photovoltaic power stations, including hourly power generation forecast data for the next 24 hours.

[0077] By integrating the above real-time operating parameters and prediction data, the digital twin model can be updated and calibrated in real time to ensure model accuracy. For example, by comparing the extraction parameters output by the model with the actual measured values, the thermodynamic parameters and flow resistance coefficient in the model can be dynamically adjusted; by comparing the actual charging and discharging efficiency of the energy storage system with the model calculated values, the loss model of the energy storage unit can be optimized; by analyzing the temperature distribution deviation of the thermal energy storage system, the thermal conductivity coefficient and convection heat transfer coefficient can be adjusted.

[0078] Based on the high-precision digital twin model, the system can perform real-time state estimation and predictive analysis. By comparing the current system state with historical data, combined with future load forecasts and renewable energy generation forecasts, the system can predict the optimal operation strategy for the next 24 hours. This includes determining the dynamic regulation curves of steam extraction flow at all levels, the charging and discharging plan of the energy storage system, and the charging and discharging strategy of the thermal energy storage system.

[0079] For example, when it is predicted that the grid load will peak in the next 4 hours, the system can increase the charging power of the compressed air energy storage system in advance, while appropriately reducing the charging rate of the phase change material thermal storage unit to prepare for the upcoming peak electricity demand. When it is predicted that the renewable energy generation is high during the late night load valley, the system can make a plan to increase the charging power of the flywheel energy storage system and increase the charging flow of the molten salt thermal storage unit to make full use of the valley power.

[0080] In this embodiment, accurate monitoring of the operating status is ensured by multi-level real-time parameter collection of the steam turbine, energy storage and thermal energy storage system; the steam extraction flow regulation, energy storage charging and discharging and thermal energy storage strategies are optimized by combining information such as grid load forecast and renewable energy power generation forecast; the accuracy and efficiency of the system are improved and energy loss is reduced by dynamically adjusting model parameters; based on future load and power generation forecasts, optimized charging and discharging and thermal energy utilization strategies are formulated in advance to improve the overall energy efficiency of the system.

[0081] In an optional embodiment, a digital twin model of an integrated system of steam extraction energy storage and thermal energy storage is established, wherein the digital twin model includes a steam turbine steam extraction system, an energy storage system, and a thermal energy storage system; the steam turbine steam extraction system includes a multi-stage steam extraction point model and a steam extraction pipeline model; the energy storage system includes a compressed air energy storage unit model and a flywheel energy storage unit model; the thermal energy storage system includes a phase change material heat storage unit model and a molten salt heat storage unit model, including:

[0082] A digital twin model of a steam turbine extraction system is established, wherein the digital twin model includes a multi-stage extraction point model and an extraction pipeline model; the multi-stage extraction point model determines the relationship between the pressure and flow rate between the steam turbine stages based on the Stodola ellipse law, and calculates the steam parameters of each stage of extraction points in combination with the heat balance equation; the extraction pipeline model is based on the one-dimensional unsteady flow theory, combined with pipeline friction, heat loss and pressure fluctuation factors, and solves the steam flow state in the pipe through the characteristic line method;

[0083] Establish a digital twin model of an energy storage system, wherein the digital twin model of the energy storage system includes a compressed air energy storage unit model and a flywheel energy storage unit model; the compressed air energy storage unit model is based on the isentropic compression-expansion theory, combined with the actual efficiency curves of the compressor and the turbine, simulates the energy conversion efficiency of the charging and discharging process, and determines the state change of the compressed air through the non-ideal gas state equation; the flywheel energy storage unit model is based on the law of conservation of angular momentum, combined with the dynamics model of the magnetic bearing system, to construct the acceleration process and deceleration process of the flywheel and determine the energy loss;

[0084] A digital twin model of a thermal energy storage system is established, wherein the digital twin model of the thermal energy storage system includes a phase change material heat storage unit model and a molten salt heat storage unit model; the phase change material heat storage unit model adopts the enthalpy-porosity method, combined with the effective thermal conductivity, to simulate the heat transfer and phase change interface movement during the phase change process; the molten salt heat storage unit model is based on the computational fluid dynamics theory, adopts the three-dimensional non-steady-state heat conduction equation, and combines the influence of natural convection and thermal radiation to determine the molten salt temperature field distribution.

[0085] The Stodola ellipse law specifically refers to the relationship between pressure and flow between each stage of a steam turbine. Assuming that at a fixed speed, the flow rate and inter-stage pressure difference of the steam turbine are distributed in an elliptical shape. The Stodola ellipse law takes into account the compressibility of gas flow and the complex flow inside the turbine, and is particularly suitable for steam turbines with multi-stage steam extraction. Through the heat balance equation, the Stodola law can accurately calculate the steam parameters (such as temperature, pressure and flow) at different steam extraction points of the steam turbine, helping to optimize the regulation and control of steam extraction.

[0086] The one-dimensional unsteady flow theory specifically refers to a theory that describes the change of fluid in a pipeline over time, and is used to analyze the steam flow in the steam extraction pipeline of a steam turbine. The one-dimensional unsteady flow theory is based on the one-dimensional unsteady flow equation, combined with factors such as friction loss, heat loss, and pressure fluctuations in the pipeline. By solving the characteristic line method, the dynamic changes in the steam flow state can be simulated, especially when the steam pipeline is started, stopped, adjusted, or suddenly changed, the transient response of pressure and flow can be accurately predicted.

[0087] The isentropic compression-expansion theory specifically means that under ideal conditions, the compression and expansion processes are isentropic, that is, no heat transfer occurs during the process (entropy remains unchanged). It is used to describe the inflation and deflation process of the compressed air energy storage unit, and simulates the energy conversion process by combining the actual efficiency curves of the compressor and expander. Through the non-ideal gas state equation, the state changes of compressed air under different pressure and temperature conditions can be described, thereby accurately estimating the efficiency of energy storage and release.

[0088] The enthalpy-porosity method specifically refers to describing the heat storage process of phase change materials. Phase change materials undergo solid-liquid or liquid-solid phase change processes when absorbing or releasing heat. The enthalpy-porosity method describes the energy state change of phase change materials by introducing enthalpy functions, and describes the movement process of phase change interfaces by combining porosity changes. It can handle complex nonlinear thermophysical properties and latent heat problems of phase change, thereby accurately simulating the heat transfer behavior of phase change materials during heat storage and heat release.

[0089] The three-dimensional non-steady-state heat conduction equation specifically refers to the description of the temperature change of an object in three-dimensional space and time dimensions. Heat conduction in the molten salt heat storage system involves complex convection, heat conduction and radiation processes. The three-dimensional non-steady-state heat conduction equation combines natural convection and thermal radiation effects to accurately simulate the changes in the temperature field in the molten salt tank, especially when the temperature gradient is large within the working temperature range, the effects of convection and radiation are significant. The equation can effectively optimize the design of the thermal energy storage system, reduce heat loss and improve thermal efficiency.

[0090] First, a digital twin model of the steam turbine extraction system is established. The model consists of two main parts: a multi-stage extraction point model and an extraction pipeline model. For the multi-stage extraction point model, the Stodola ellipse law is used to describe the relationship between the pressure and flow rate between the turbine stages. Taking a 300MW steam turbine as an example, it can be divided into three main extraction points: high pressure, medium pressure and low pressure. For the high-pressure extraction point, it is assumed that the steam parameters under rated conditions are 16.7MPa, 538℃, and the flow rate is 950t / h. Using the Stodola ellipse law, a functional relationship between the turbine inlet pressure, extraction pressure and flow rate can be established. For example, when the unit load drops to 80%, the relationship can be used to calculate that the pressure at the high-pressure extraction point drops to about 3.8MPa and the flow rate is reduced to about 60t / h.

[0091] Combined with the heat balance equation, the steam temperature and enthalpy of each extraction point can be further calculated. For example, for the medium-pressure extraction point, under 80% load conditions, it can be calculated that its temperature is about 320°C and its enthalpy is 3050 kJ / kg. These parameters will be used as input conditions for the subsequent extraction pipeline model.

[0092] The steam extraction pipeline model is established based on the one-dimensional unsteady flow theory. Taking the high-pressure steam extraction pipeline as an example, it is assumed that the pipeline length is 50m and the inner diameter is 0.5m. Considering the factors of pipeline friction, heat loss and pressure fluctuation, the characteristic line method is used to solve the steam flow state in the pipe. By dividing the pipeline into multiple microelements and discretizing them in time and space, the pressure, temperature and flow velocity distribution at any position and time in the pipeline can be obtained. For example, under rated conditions, it can be calculated that the pressure loss at the end of the pipeline is about 0.2MPa and the temperature drops by about 5°C.

[0093] Next, a digital twin model of the energy storage system is established, including a compressed air energy storage unit model and a flywheel energy storage unit model. For the compressed air energy storage unit model, the isentropic compression-expansion theory is used and the efficiency curves of the actual compressor and turbine are used for correction. Taking a 100MW compressed air energy storage system as an example, assuming that the air storage tank volume is 50,000m 3 , the working pressure range is 4-8MPa. During the inflation process, the multi-stage compression process of the compressor can be simulated, and the influence of inter-stage cooling can be considered. For example, under the condition of an ambient temperature of 25°C, after four stages of compression, the air temperature may rise to about 150°C, and the temperature needs to be reduced to below 50°C by a cooler before entering the air storage tank.

[0094] Through the non-ideal gas state equation, the state changes of compressed air at different pressures and temperatures can be accurately described. For example, when the pressure of the gas tank rises from 4MPa to 8MPa, it can be calculated that the stored energy is about 90MWh. During the deflation process, considering the actual efficiency curve of the turbine, the output power and efficiency under different working conditions can be simulated. For example, at an initial pressure of 6MPa, the system may achieve an overall efficiency of 65% and an output power of 60MW.

[0095] The flywheel energy storage unit model is established based on the law of conservation of angular momentum. Taking a 10MWh flywheel energy storage system as an example, it is assumed that the flywheel rotor mass is 5 tons and the maximum speed is 16000rpm. By establishing a dynamic model of the magnetic bearing system, the acceleration and deceleration process of the flywheel can be accurately described. For example, during the charging process, it can be simulated that the time required for the flywheel to accelerate from rest to 16000rpm is about 15 minutes, and the charging efficiency after considering various losses is about 95%.

[0096] When determining energy losses, factors such as bearing friction, air resistance, and motor / generator losses need to be considered. By building a detailed loss model, the self-discharge rate of the flywheel at different speeds can be predicted. For example, at full load, the 24-hour self-discharge rate of the system may be around 2%.

[0097] Finally, a digital twin model of the thermal energy storage system is established, including a phase change material heat storage unit model and a molten salt heat storage unit model. The phase change material heat storage unit model adopts the enthalpy-porosity method, combined with the concept of effective thermal conductivity. Taking the 200MWh phase change material heat storage unit as an example, it is assumed that paraffin is used as the phase change material and the phase change temperature range is 58-62°C. By dividing the heat storage device into fine grids, the heat transfer and phase change interface movement during the phase change process can be simulated. Exemplarily, during the charging process, it can be calculated that the time required for 90% of the material to complete the phase change from the beginning is about 4 hours, and the thermal storage efficiency reaches 92%.

[0098] The molten salt thermal storage unit model is based on computational fluid dynamics theory. Taking a 300MWh molten salt thermal storage unit as an example, assuming that a mixed salt of 60% NaNO3 and 40% KNO3 is used, the operating temperature range is 290-565°C. Using the three-dimensional non-steady-state heat conduction equation and considering the effects of natural convection and thermal radiation, the temperature field distribution in the molten salt tank can be accurately simulated. For example, during the charging process, the temperature stratification phenomenon generated in the tank can be calculated, and the temperature difference between the top and the bottom may reach 50°C. By optimizing the heat exchanger design and the molten salt flow path, the uniformity of temperature distribution can be improved and the heat storage efficiency can be improved.

[0099] Through the above steps, a complete digital twin model of the integrated system of extraction steam storage and thermal energy storage can be established. This model can accurately reflect the dynamic characteristics and interactions of each subsystem, and provide strong support for the optimization operation and control strategy formulation of the system. For example, based on this model, simulation comparisons of different operation schemes can be carried out to evaluate the system response capabilities under conditions such as grid load fluctuations and changes in the proportion of renewable energy access. Through digital twin technology, predictive maintenance and fault diagnosis of the system can be achieved, improving equipment reliability and overall system efficiency.

[0100] In this embodiment, the dynamic characteristics of the multi-stage steam extraction system of the steam turbine are accurately simulated by Stodola's ellipse law and one-dimensional unsteady flow theory, the steam extraction flow control is optimized, the efficiency of the steam turbine is improved, and the energy loss is reduced; the isentropic compression-expansion theory and the flywheel energy storage model are used to accurately describe the energy conversion and loss of compressed air and flywheel energy storage units, optimize the charging and discharging process, and significantly improve the efficiency of the energy storage system; the enthalpy-porosity method and the three-dimensional unsteady heat conduction equation are used to accurately simulate the heat conduction and phase change process of the phase change material and the molten salt heat storage unit, reduce heat loss, and ensure that the thermal energy storage system operates in a high-efficiency state; combined with load forecasting and renewable energy power generation forecasting, steam turbine extraction, energy storage charging and discharging, and thermal energy storage strategies are formulated in advance to improve the system's responsiveness and overall economy; through real-time data acquisition and model self-calibration, the digital twin model is ensured to be always accurate, and the system can dynamically adjust the operation strategy to further improve energy efficiency and reliability.

[0101] S102. Establish a multi-objective optimization model based on the digital twin model of the integrated system of extraction steam storage and thermal energy storage, the first real-time operating parameter, the second real-time operating parameter, the third real-time operating parameter, the real-time load curve of the power grid, the load forecast data for the next 24 hours, the ambient temperature change trend and the renewable energy power generation forecast information; use the improved multi-objective differential evolution algorithm, combined with the adaptive mutation strategy and the non-dominated sorting mechanism, to solve the multi-objective optimization model and obtain the optimal adjustment strategy;

[0102] First, collect the real-time data and forecast information required for system operation. The first real-time operating parameter includes the steam extraction pressure, temperature and flow rate of each level of the steam turbine extraction system. Taking the 300MW unit as an example, the real-time parameters of the high-pressure extraction point may be: pressure 3.8MPa, temperature 380℃, flow rate 60t / h. The second real-time operating parameter covers the state of the energy storage system, such as the pressure and temperature of the gas tank of the compressed air energy storage unit, the speed of the flywheel energy storage unit, etc. Exemplarily, the gas tank pressure of the compressed air energy storage unit is 6.5MPa and the temperature is 40℃; the current speed of the flywheel energy storage unit is 12000rpm. The third real-time operating parameter includes the operating state of the thermal energy storage system, such as the temperature distribution of the phase change material, the liquid level and temperature of the molten salt heat storage tank, etc. For example, the average temperature of the phase change material is 60℃, the liquid level of the molten salt heat storage tank is 80%, the top temperature is 540℃, and the bottom temperature is 500℃.

[0103] Next, obtain the real-time load curve of the power grid and the load forecast data for the next 24 hours. The real-time load curve of the power grid reflects the current power demand. For example, the current load is 250MW, showing a slow upward trend. The load forecast data for the next 24 hours may show that there will be two peaks of power consumption at 7:00-9:00 in the morning and 18:00-20:00 in the evening the next day, with peak loads reaching 280MW and 290MW respectively.

[0104] At the same time, the environmental temperature change trend and renewable energy power generation forecast information are collected. The environmental temperature change trend has an important impact on system efficiency and thermal energy storage. It is predicted that the temperature will fluctuate between 15-28°C in the next 24 hours, reaching the highest temperature at 12:00-14:00 noon. The renewable energy power generation forecast information includes the output forecast of photovoltaic and wind power. For example, it is expected that the photovoltaic power generation will reach its peak at 11:00-15:00 the next day, with a maximum output of 50MW; wind power is relatively abundant at night from 22:00 to 4:00 the next morning, with a maximum output of 30MW.

[0105] Based on the above data and the digital twin model of the integrated system of extraction steam energy storage and thermal energy storage, a multi-objective optimization model is established. The objective functions of the model include: maximizing the overall efficiency of the system, minimizing the operating cost, and maximizing the proportion of renewable energy consumption. Constraints include: equipment operating parameter restrictions, grid balance requirements, energy storage capacity restrictions, etc.

[0106] Taking the overall efficiency of the system as an example, the energy conversion efficiency under different working conditions can be calculated through the digital twin model. For example, under the current working conditions, the charging and discharging efficiency of the compressed air energy storage unit is 65%, the round-trip efficiency of the flywheel energy storage unit is 90%, the thermal storage efficiency of the phase change material heat storage unit is 92%, and the thermal storage efficiency of the molten salt heat storage unit is 95%. By adjusting the operating parameters of each subsystem, the overall efficiency can be optimized.

[0107] Operating costs include equipment wear costs, electricity purchase costs, and maintenance costs. For example, the wear of a flywheel energy storage unit is related to the number and depth of starts and stops, and its life can be extended by controlling its operating range; the operating cost of a compressed air energy storage unit is related to the efficiency curve of the compressor and turbine, and the optimal operating point can be selected.

[0108] The renewable energy consumption ratio reflects the system's ability to absorb intermittent energy. By coordinating the operation of energy storage and thermal storage systems, it is possible to increase storage during peak renewable energy generation periods and release during off-peak periods, thereby increasing the consumption ratio.

[0109] In order to solve the multi-objective optimization model, an improved multi-objective differential evolution algorithm is used. The core idea of ​​the algorithm is to continuously iterate and optimize the solution set by simulating the population evolution process. On the basis of the standard differential evolution algorithm, an adaptive mutation strategy and a non-dominated sorting mechanism are introduced to improve the convergence speed of the algorithm and the diversity of solutions.

[0110] The adaptive mutation strategy dynamically adjusts the mutation parameters according to the evolutionary state of the population. For example, in the early stage of optimization, a larger mutation rate (such as 0.8) can be set to increase the solution exploration ability; as the iteration proceeds, the mutation rate is gradually reduced (such as to 0.2) to enhance the local search ability. In this way, a balance can be achieved between the global optimum and the local optimum.

[0111] The non-dominated sorting mechanism is used to evaluate the quality of solutions and guide the direction of population evolution. In specific implementation, all solutions can be sorted according to the Pareto principle, and non-dominated solutions are retained first. For example, 100 candidate solutions are obtained in one iteration, and 20 non-dominated solutions may be screened out through non-dominated sorting to form the Pareto frontier. These solutions have their own advantages in different objectives, providing decision makers with a variety of choices.

[0112] During the execution of the algorithm, the digital twin model needs to be called repeatedly for simulation calculations. For example, for each candidate solution, the digital twin model needs to be used to evaluate its operating effect in the next 24 hours, including system efficiency, operating costs, and renewable energy consumption in each period. This process may need to be repeated thousands of times until the convergence condition is met or the maximum number of iterations is reached.

[0113] Finally, the algorithm will output a set of non-dominated solutions, that is, the optimal regulation strategy set. Each strategy contains the operating parameter curves of each subsystem in the next 24 hours. For example, the compressed air energy storage unit is charged at maximum power during the peak period of renewable energy generation (such as 11:00-15:00), and the pressure of the gas tank increases from 5MPa to 8MPa; during the peak period of electricity consumption (such as 18:00-20:00), it is discharged at maximum power, and the pressure drops to 5.5MPa. The flywheel energy storage unit is mainly used for fast response, providing instantaneous power support when the load changes rapidly (such as 17:30-18:00), and the speed fluctuates between 10000-15000rpm. The thermal energy storage system coordinates operation according to the heat load demand and power generation demand. For example, during the low period at night (23:00-5:00 the next day), low-priced electricity is used to heat the molten salt heat storage tank, and the temperature rises from 400℃ to 550℃, providing a heat source for the peak period during the day.

[0114] This set of optimal regulation strategies provides a variety of options for system operation, and operators can choose the most appropriate strategy to execute according to actual needs and preferences. By implementing these optimization strategies, the system's operating efficiency, economy and flexibility can be significantly improved, providing strong support for the safe and stable operation of the power grid and the large-scale access of renewable energy.

[0115] In this embodiment, by collecting real-time data and load forecasts, using multi-objective optimization models and digital twin technology, the operating parameters of each subsystem are dynamically adjusted to significantly improve the overall efficiency of the system; the working points of equipment such as flywheel energy storage and compressed air energy storage are optimized to extend equipment life and reduce electricity purchase and maintenance costs; by coordinating the energy storage system and the thermal energy storage system, energy is stored during peak renewable energy generation and released during low periods, thereby increasing the utilization ratio of renewable energy and improving the system's ability to absorb intermittent energy; an improved multi-objective differential evolution algorithm is used to output a variety of optimal adjustment plans, achieve flexible scheduling decisions, and enhance the adaptability and flexibility of the system; by optimizing the system's operating strategy, support is provided for the safe operation of the power grid and the access of renewable energy, thereby improving the stability and sustainability of the power system.

[0116] In an optional embodiment, the multi-objective optimization model is solved by using an improved multi-objective differential evolution algorithm, combined with an adaptive mutation strategy and a non-dominated sorting mechanism, to obtain the optimal adjustment strategy, including:

[0117] Set the population size and maximum number of iterations, randomly generate an initial feasible solution set that meets the constraints as the initial population, use the initial population as the current population, and start iteration:

[0118] Performing non-dominated sorting on the current population, dividing the individuals in the current population into different non-dominated levels, calculating the crowding distance of individuals in the same non-dominated level in the current population, sorting the current population based on the non-dominated level and the crowding distance to obtain an ordered population, calculating the fitness improvement rate of the ordered population, and dynamically adjusting the mutation factor based on the fitness improvement rate;

[0119] Use the adjusted mutation factor to perform mutation operation on each individual in the ordered population, randomly select three different individuals, generate mutant individuals through linear combination to obtain a mutant population, perform binomial crossover between the mutant population and the ordered population to generate a crossover population, and evaluate the objective function value of each individual in the crossover population based on the multi-objective optimization model;

[0120] Merge the ordered population and the crossover population, perform non-dominated sorting and crowding distance calculation on the merged population again, obtain new non-dominated sorting results and new crowding distances, and select the best individuals from the merged population to form a new generation of population based on the new non-dominated sorting results and new crowding distances;

[0121] The new generation population is taken as the current population, and the iteration is repeated until the preset maximum number of iterations is reached. The non-dominated solution set is extracted from the last generation population to determine the Pareto optimal solution set. The fuzzy comprehensive evaluation method is used to combine the decision maker's preferences and actual operating conditions to select the final optimal solution from the Pareto optimal solution set.

[0122] First, set the key parameters of the algorithm. The population size is usually selected between 50 and 200, and is set to 100 in this embodiment. The maximum number of iterations depends on the complexity of the problem and the required accuracy, and is set to 1000 here.

[0123] Next, an initial set of feasible solutions that meet the constraints is randomly generated as the initial population. Each solution represents a possible system operation strategy, which contains the operating parameters of each subsystem in the next 24 hours. For example, for compressed air energy storage units, the charging and discharging power per hour within 24 hours can be randomly generated, ranging from -50MW (discharging) to 50MW (charging); for thermal energy storage systems, the heat storage and withdrawal per hour within 24 hours can be randomly generated, ranging from -100MWh (heat release) to 100MWh (heat storage). Various constraints need to be considered during the generation process, such as equipment operating parameter restrictions, grid balance requirements, energy storage capacity restrictions, etc.

[0124] The iteration process starts with the initial population as the current population. The first step is to perform non-dominated sorting on the current population. The purpose of non-dominated sorting is to divide the individuals in the population into different non-dominated levels. For the multi-objective optimization problem of this system, the objective function includes the overall efficiency of the system, the operating cost, and the proportion of renewable energy consumption. The non-dominated sorting process is as follows: first find the solutions that are not dominated by any other individuals in the population and classify them as the first non-dominated level; then, find non-dominated solutions again in the remaining solutions and classify them as the second non-dominated level; repeat this process until all solutions are assigned levels.

[0125] For example, suppose there are solution A (90% efficiency, 10 million yuan cost, 80% renewable energy consumption rate) and solution B (88% efficiency, 9.5 million yuan cost, 82% renewable energy consumption rate). Since A is better than B in efficiency, and B is better than A in cost and renewable energy consumption rate, they do not dominate each other and may belong to the first non-dominated level.

[0126] Next, the crowding distance of individuals in the same non-dominated level is calculated. The crowding distance reflects the distribution of solutions in the target space and is used to maintain the diversity of the population. The calculation method is that for each objective function, the solutions of the same level are sorted according to the target value, and then the normalized distance between adjacent solutions is calculated. The crowding distance of the boundary solution (that is, the solution that achieves the maximum or minimum value on a certain objective) is set to infinity.

[0127] For example, for the efficiency goal, if there are solution C (efficiency 89%) and solution D (efficiency 91%) in the same level, the normalized distance between them may be 0.02 (assuming that the global maximum and minimum values ​​of efficiency are 95% and 85% respectively).

[0128] The current population is sorted based on the non-dominated level and crowding distance to obtain an ordered population. The sorting rule is: first sort by non-dominated level from low to high, and if the levels are the same, sort by crowding distance from large to small. In this way, individuals with low non-dominated level (i.e. better) and large crowding distance (i.e. more isolated in the target space) can be retained first.

[0129] Subsequently, the fitness improvement rate of the ordered population is calculated. The fitness improvement rate reflects the speed of population quality improvement and can be used to dynamically adjust the mutation factor. The calculation method is to compare the improvement of the optimal solution of the current generation with that of the previous generations in each objective function. For example, if the optimal solution of the current generation improves efficiency, cost and renewable energy consumption rate by 0.5%, decreases by 1% and increases by 0.8% respectively compared with the previous generation, then the fitness can be considered to have improved.

[0130] Dynamically adjust the mutation factor based on the fitness improvement rate. When the fitness improvement rate is high, it means that the population still has a large optimization space, and the mutation factor can be increased to enhance the exploration ability; when the fitness improvement rate is low, it means that the population is close to the optimal solution, and the mutation factor can be reduced to enhance the development ability. For example, the initial value of the mutation factor can be set to 0.5. When the fitness improvement rate is higher than 5%, the mutation factor is increased by 0.1; when the fitness improvement rate is lower than 1%, the mutation factor is reduced by 0.1, but not lower than 0.1.

[0131] Use the adjusted mutation factor to perform mutation operations on each individual in the ordered population. The purpose of the mutation operation is to generate new candidate solutions and increase the diversity of the population. The specific steps are: for each individual to be mutated, three different individuals are randomly selected and the mutant individuals are generated through linear combination. The weight of the linear combination is related to the mutation factor. For example, the weight can be set to the mutation factor, 1-mutation factor, and 0.

[0132] For example, assuming that the compressed air energy storage charging powers of the three selected individuals at a certain moment are 20MW, 30MW and 40MW respectively, and the variation factor is 0.6, then the charging power after variation may be 20*0.6+30*(1-0.6)+40*0=24MW. Similar operations are performed on each decision variable, and finally a complete variation individual is obtained.

[0133] After all individuals are mutated, a mutant population is obtained. The mutant population is subjected to binomial crossover with the ordered population to generate a crossover population. The process of binomial crossover is: for each decision variable, a certain probability (called the crossover rate, usually set to 0.5) is used to decide whether to crossover. If crossover is performed, the value of the variable is inherited from the mutant individual; otherwise, the value in the original ordered population remains unchanged.

[0134] For example, for the heat storage and access amount of the thermal energy storage system at a certain moment, if the randomly generated probability value is less than the crossover rate, the value in the mutant population is adopted, otherwise the value in the ordered population is maintained. In this way, new changes can be introduced while retaining the original excellent features.

[0135] Next, the objective function value of each individual in the cross population based on the multi-objective optimization model is evaluated. This requires calling the digital twin model of the integrated system of extraction steam storage and thermal energy storage to simulate the performance of the operating strategy represented by each individual in the next 24 hours. The evaluation indicators include the overall efficiency of the system, operating costs and the proportion of renewable energy consumption.

[0136] For example, for a certain individual, the digital twin model calculates that the 24-hour average system efficiency is 88%, the total operating cost is 9.8 million yuan, and the renewable energy consumption rate is 85%. These values ​​will be used in the subsequent non-dominated sorting and selection process.

[0137] The ordered population and the crossover population are merged, and the non-dominated sorting and crowding distance calculations are performed again on the merged population to obtain a new non-dominated sorting result and a new crowding distance. This process is the same as described before, but it deals with a larger population (twice the size of the original population).

[0138] Based on the new non-dominated sorting results and the new crowding distance, the best individuals are selected from the merged population to form a new generation of population. The selection process follows the elite retention strategy, that is, individuals with low non-dominated levels are given priority, and individuals with large crowding distances are given priority within the same level, until the size of the new population reaches the preset value (the same as the original population).

[0139] The new generation population thus obtained is used as the current population for the next round of iteration, and the above process is repeated until the preset maximum number of iterations is reached. In this embodiment, this process will be repeated 1000 times.

[0140] When the iteration is over, the non-dominated solution set is extracted from the last generation of population to determine the Pareto optimal solution set. Each solution in the Pareto optimal solution set represents the optimal trade-off solution for certain objectives.

[0141] Finally, the fuzzy comprehensive evaluation method is used to select the final optimal solution from the Pareto optimal solution set, combining the decision maker's preferences and actual operating conditions. The fuzzy comprehensive evaluation method takes into account the importance of multiple objectives and the subjective preferences of decision makers. For example, the weights of system efficiency, operating cost, and renewable energy consumption rate can be set to 0.4, 0.3, and 0.3, respectively. The performance of each solution on these three objectives is fuzzily evaluated to obtain a comprehensive score. The solution with the highest comprehensive score is selected as the final optimal solution.

[0142] In this embodiment, the operating parameters of each sub-module of the system are dynamically adjusted through the differential evolution algorithm to improve the overall operating efficiency, reduce costs, and increase the proportion of renewable energy consumption; a multi-objective optimization model is used to balance multiple indicators such as system efficiency, operating costs and renewable energy consumption, and finally a set of Pareto optimal solutions is obtained, providing multiple operating strategies for selection; the diversity of the population is maintained through non-dominated sorting and congestion distance, and the mutation factor is dynamically adjusted to increase the convergence speed and improve the optimization efficiency; finally, the fuzzy comprehensive evaluation method is combined with the decision maker's preferences to select the best solution from the Pareto optimal solution set to ensure that the optimization results meet the actual operation needs.

[0143] S103. Based on the optimal regulation strategy, a multi-stage steam extraction flow regulation instruction is sent to the steam turbine extraction system through the distributed control system, and the fuzzy PID control algorithm is used to adjust the opening of the steam extraction valves at each level of extraction points in real time to control the steam extraction flow; a coordinated charging and discharging instruction is sent to the energy storage system, and the charging and discharging power of the compressed air energy storage unit is dynamically adjusted according to the load changes of the power grid and the electricity price information, and the charging and discharging power of the flywheel energy storage unit is controlled at the same time; a coordinated heat storage instruction is sent to the thermal energy storage system, and the charging and discharging rates of the phase change material heat storage unit are adaptively adjusted according to the heat load demand and the ambient temperature changes, and the charging and discharging flow of the molten salt heat storage unit is controlled at the same time.

[0144] In this embodiment, the opening of the steam extraction valve is adjusted in real time through the fuzzy PID control algorithm to ensure that the steam extraction flow rate of each stage of the turbine accurately matches the load demand, thereby improving the system response speed and stability; according to the changes in grid load and electricity prices, the charging and discharging strategies of compressed air energy storage and flywheel energy storage units are dynamically optimized to improve the flexibility and economic benefits of the energy storage system; the charging and discharging rates of phase change materials and molten salt heat storage units are adaptively adjusted to ensure that the thermal energy storage system operates efficiently according to demand and maximize the thermal energy utilization efficiency.

[0145] In an optional embodiment, based on the optimal regulation strategy, a multi-stage steam extraction flow regulation instruction is sent to the steam turbine extraction system through a distributed control system, and a fuzzy PID control algorithm is used to adjust the opening of the steam extraction valves at each level of extraction points in real time to control the steam extraction flow; a coordinated charging and discharging instruction is sent to the energy storage system, and the charging power and discharging power of the compressed air energy storage unit are dynamically adjusted according to the load changes of the power grid and the electricity price information, and the charging and discharging power of the flywheel energy storage unit is controlled at the same time; a coordinated heat storage instruction is sent to the thermal energy storage system, and the charging and discharging rates of the phase change material heat storage unit are adaptively adjusted according to the heat load demand and the ambient temperature changes, and the charging and discharging flow of the molten salt heat storage unit is controlled at the same time, including:

[0146] According to the multi-stage steam extraction flow dynamic regulation curve in the optimal regulation strategy, the actual steam extraction flow of each stage of steam extraction point is collected, the deviation and the deviation change rate from the target steam extraction flow are calculated, and the deviation and the deviation change rate are fuzzy processed to divide them into fuzzy subsets;

[0147] Based on the fuzzy subset, an if-then fuzzy rule base is established, and the centroid method is used for defuzzification to obtain the adjustment amount of the proportional integral differential controller parameters, and the opening adjustment amount of the steam extraction valve is calculated by updating the proportional integral differential controller parameters, and the opening adjustment amount is sent to the corresponding steam extraction valve actuator through the distributed control system;

[0148] Obtaining grid load data and electricity price information in real time, calculating the load deviation and electricity price gradient at the corresponding moment, and dividing the system operation state into multiple operation modes based on the corresponding values ​​of the load deviation and electricity price gradient, including peak-valley filling mode, peak shaving mode, valley filling mode and economic operation mode;

[0149] Based on different operation modes, dynamically adjust the first charging and discharging power of the compressed air energy storage unit and the second charging and discharging power of the flywheel energy storage unit, and send the adjustment instructions of the first charging and discharging power and the second charging and discharging power to the corresponding charging and discharging actuators through the distributed control system;

[0150] Collect heat load demand and ambient temperature in real time, calculate heat load deviation and ambient temperature change rate, and divide the operation state of the thermal energy storage system into a fast response mode, a stable operation mode and an energy-saving mode according to the corresponding values ​​of the heat load deviation and the ambient temperature change rate;

[0151] The first current heat output of the phase change material heat storage unit and the second current heat storage of the molten salt heat storage unit are calculated; according to the ratio of the first current heat storage to the second current heat storage, the charging and discharging power distribution ratio of the phase change material heat storage unit and the molten salt heat storage unit are dynamically adjusted; based on the charging and discharging power distribution ratio, the charging and discharging rate of the phase change material heat storage unit and the charging and discharging flow rate of the molten salt heat storage unit are calculated, and sent to the corresponding charging and discharging actuators through the distributed control system.

[0152] First, according to the multi-stage steam extraction flow dynamic adjustment curve in the optimal adjustment strategy, the actual steam extraction flow of each steam extraction point is collected. Taking a 300MW unit as an example, it is assumed that there are three steam extraction points of high pressure, medium pressure and low pressure, and each steam extraction point is equipped with a flow meter. The data of these flow meters are read in real time through the distributed control system to obtain the actual steam extraction flow of each steam extraction point.

[0153] Next, the deviation and deviation change rate of the actual steam extraction flow rate and the target steam extraction flow rate are calculated. For example, assuming that the target flow rate of the high-pressure steam extraction point at a certain moment is 50t / h, the actual flow rate is 48t / h, the actual flow rate at the previous moment is 47t / h, and the sampling time interval is 1 minute. Then the deviation can be calculated to be -2t / h, and the deviation change rate is 1t / h / min.

[0154] The calculated deviation and deviation change rate are fuzzified and divided into fuzzy subsets. The deviation can be divided into five fuzzy subsets: "negative large", "negative small", "zero", "positive small", and "positive large", and the deviation change rate can be divided into three fuzzy subsets: "negative", "zero", and "positive". Each fuzzy subset has a membership function that describes the degree to which the actual value belongs to the subset. For example, a deviation of -2t / h may have a membership of 0.6 belonging to the "negative small" subset and a membership of 0.4 belonging to the "negative large" subset.

[0155] Based on these fuzzy subsets, an if-then fuzzy rule base is established. The rule base contains a series of control rules summarized by expert experience, for example: "If the deviation is negative and the deviation change rate is positive, the proportional gain is slightly increased, the integral time constant is slightly decreased, and the differential time constant remains unchanged." Similar rules will cover all possible combinations of deviations and deviation change rates.

[0156] Using the established fuzzy rule base, the centroid method is used for defuzzification to obtain the adjustment amount of the proportional integral derivative (PID) controller parameters. The centroid method is a commonly used defuzzification method, which obtains the final clear output by calculating the weighted average of the outputs of each rule. For example, after defuzzification, the adjustment amount may be increased by 0.05 for the proportional gain, decreased by 0.1 seconds for the integral time constant, and increased by 0.02 seconds for the differential time constant.

[0157] By updating the PID controller parameters, the opening adjustment of the steam extraction valve is calculated. Assume that the original PID parameters are: proportional gain 1.2, integral time constant 10 seconds, differential time constant 2 seconds. The updated parameters are: proportional gain 1.25, integral time constant 9.9 seconds, differential time constant 2.02 seconds. Using the updated parameters, combined with the current deviation and deviation change rate, the valve opening adjustment is calculated, exemplarily increasing the opening by 2%.

[0158] The calculated opening adjustment amount is sent to the corresponding steam extraction valve actuator through the distributed control system. After receiving the instruction, the actuator will adjust the valve opening accordingly, thereby changing the steam extraction flow rate and gradually approaching the target value.

[0159] At the same time, the grid load data and electricity price information are obtained in real time. These data can be obtained in real time through the data interface with the grid dispatching center. For example, the grid load at a certain moment is 10000MW and the electricity price is 0.5 yuan / kWh.

[0160] Calculate the load deviation and electricity price gradient at the corresponding time. The load deviation can be obtained by comparing the current load with the load forecasted the day before, and the electricity price gradient is the rate of change of electricity prices in adjacent time periods. For example, if the load forecasted the day before is 9800MW, the load deviation is 200MW; if the electricity price in the next period is 0.6 yuan / kWh, the electricity price gradient is 0.1 yuan / kWh / h.

[0161] Based on the calculated corresponding values ​​of load deviation and electricity price gradient, the system operation state is divided into multiple operation modes. The following rules can be set: when the load deviation is within the range of ±100MW and the absolute value of the electricity price gradient is less than 0.05 yuan / kWh / h, the peak-valley filling mode is adopted; when the load deviation is greater than 200MW, or the electricity price gradient is greater than 0.1 yuan / kWh / h, the peak shaving mode is adopted; when the load deviation is less than -200MW, or the electricity price gradient is less than -0.1 yuan / kWh / h, the valley filling mode is adopted; in other cases, the economic operation mode is adopted.

[0162] According to the determined operation mode, the first charging and discharging power of the compressed air energy storage unit and the second charging and discharging power of the flywheel energy storage unit are dynamically adjusted. For example, in the peak shaving mode, the compressed air energy storage unit can be set to discharge and generate electricity at a power of 40MW, and the flywheel energy storage unit can be set to discharge at a power of 10MW. In the valley filling mode, the compressed air energy storage unit can be set to charge and store energy at a power of 30MW, and the flywheel energy storage unit can be set to charge at a power of 15MW.

[0163] The regulation instructions of the first charging and discharging power and the second charging and discharging power are sent to the corresponding charging and discharging actuators through the distributed control system. These actuators include the compressor and expander of the compressed air energy storage system, and the motor / generator of the flywheel energy storage system.

[0164] At the same time, the heat load demand and ambient temperature are collected in real time. The heat load demand can be obtained through the heat metering device of the heat user, and the ambient temperature is measured by the temperature sensor. For example, the heat load demand at a certain moment is 200MW and the ambient temperature is 15°C.

[0165] Calculate the heat load deviation and ambient temperature change rate. The heat load deviation can be obtained by comparing the actual heat load with the predicted heat load, and the ambient temperature change rate is the change in temperature per unit time. For example, if the predicted heat load is 190MW, the heat load deviation is 10MW; if the ambient temperature one hour ago was 13°C, the ambient temperature change rate is 2°C / h.

[0166] According to the calculated corresponding values ​​of heat load deviation and ambient temperature change rate, the operation state of the thermal energy storage system is divided into fast response mode, stable operation mode and energy-saving mode. The following rules can be set: when the absolute value of the heat load deviation is greater than 20MW, or the absolute value of the ambient temperature change rate is greater than 5℃ / h, the fast response mode is adopted; when the absolute value of the heat load deviation is between 5-20MW, and the absolute value of the ambient temperature change rate is between 1-5℃ / h, the stable operation mode is adopted; when the absolute value of the heat load deviation is less than 5MW, and the absolute value of the ambient temperature change rate is less than 1℃ / h, the energy-saving mode is adopted.

[0167] Next, the first current heat storage capacity of the phase change material heat storage unit and the second current heat storage capacity of the molten salt heat storage unit are calculated. This can be calculated by measuring the temperature and flow rate of the heat storage medium and combining the thermophysical parameters of the heat storage material. For example, it is assumed that the phase change material heat storage unit currently stores 500MWh of heat and the molten salt heat storage unit stores 1000MWh of heat.

[0168] According to the ratio of the first current heat storage amount to the second current heat storage amount, the heat charging and discharging power distribution ratio of the phase change material heat storage unit and the molten salt heat storage unit is dynamically adjusted. In this example, the heat storage ratio of the phase change material heat storage unit to the molten salt heat storage unit is 1:2. According to this ratio, the distribution of the heat charging and discharging power can be determined in combination with the characteristics of different operating modes. For example, in the fast response mode, 70% of the heat charging and discharging power can be allocated to the phase change material heat storage unit with a faster response speed, and 30% to the molten salt heat storage unit.

[0169] Based on the determined charging and discharging power distribution ratio, the charging and discharging rate of the phase change material heat storage unit and the charging and discharging flow rate of the molten salt heat storage unit are calculated. Assuming that the total charging power demand is 100MW, the charging rate of the phase change material heat storage unit is 70MW, and the charging flow rate of the molten salt heat storage unit can be calculated based on the specific heat capacity and temperature difference of the molten salt, exemplarily 500t / h.

[0170] Finally, the calculated heat charging and discharging rate and heat charging and discharging flow are sent to the corresponding heat charging and discharging actuators through the distributed control system. These actuators include the heat exchanger control valve of the phase change material heat storage unit, and the circulation pump and control valve of the molten salt heat storage unit.

[0171] In this embodiment, the valve opening of each level of steam extraction point is dynamically adjusted to accurately control the steam extraction flow rate to match the target value, thereby improving the stability and efficiency of the unit operation; based on the changes in grid load and electricity price, the charging and discharging power of compressed air and flywheel energy storage units are intelligently adjusted to achieve peak shaving and valley filling and improve economic benefits; according to the heat load demand and ambient temperature changes, the charging and discharging power distribution of phase change materials and molten salt heat storage units is dynamically adjusted to ensure the rapid response and stable operation of the thermal energy storage system

[0172] In an optional embodiment, the charging and discharging heat power distribution ratio is expressed as follows:

[0173]

[0174] Among them, α represents the charging and discharging heat power allocation ratio, k 1 represents the first weight, S PCM represents the current heat storage capacity of the phase change material heat storage unit, S salt represents the current heat storage capacity of the molten salt heat storage unit, k 2represents the second weight, S salt-max Indicates the maximum heat storage capacity of the molten salt heat storage unit;

[0175] The heat charging and discharging rate of the phase change material heat storage unit is as follows:

[0176]

[0177] Among them, R PCM represents the charging and discharging rate of the phase change material heat storage unit, ΔQ represents the heat load deviation, and ΔH PCM Represents the latent heat of phase change of phase change material;

[0178] The charging and discharging heat flow of the molten salt heat storage unit is as follows:

[0179]

[0180] Among them, F salt represents the heat charging and discharging flow rate of the molten salt heat storage unit, c p-salt represents the specific heat capacity of the molten salt, ΔT salt Represents the temperature difference of the molten salt.

[0181] The formula for the power distribution ratio of heat charging and discharging dynamically adjusts the power distribution of the phase change material and molten salt heat storage units by combining their current heat storage. The formula comprehensively considers the current status of the two heat storage units to ensure that the phase change material unit and the molten salt unit play the best role in the heat charging and discharging process. By balancing the heat storage of the two, the overall efficiency of the system is optimized.

[0182] The heat charging and discharging rate formula of the phase change material heat storage unit calculates the actual heat charging and discharging rate of the phase change material heat storage unit, and ensures that the heat storage rate matches the actual demand based on the power allocation ratio and the heat load deviation, thereby enhancing the rapid response capability of the system.

[0183] The charging and discharging flow rate formula of the molten salt heat storage unit determines the charging and discharging flow rate of the molten salt heat storage unit, and reasonably regulates the flow rate by allocating power and temperature changes, thereby achieving efficient operation of the heat storage unit under different working conditions.

[0184] In an optional embodiment, the method further includes:

[0185] Setting up multiple sensors, acquiring real-time data from the sensors, performing data cleaning and preprocessing on the real-time data from the sensors to obtain real-time operation data, inputting the real-time operation data into a deep learning model, performing feature extraction on the real-time operation data, outputting the predicted status of each component of the system, and determining the abnormal situation prediction result;

[0186] Based on the abnormal situation prediction results, the cumulative deviation of multiple control parameters is calculated. When the cumulative deviation exceeds the preset control threshold, it is determined that the system operation state has undergone an over-threshold change, triggering the online adaptive optimization mechanism, resetting all control parameters to safe default values, and increasing the sampling frequency of multiple sensors to obtain high-frequency sampling data;

[0187] The digital twin model is updated based on the high-frequency sampling data using the recursive least squares method, and based on the current state of the model and external environment information, the optimization objective function is redefined, the optimal adjustment strategy is re-determined, and the optimal adjustment strategy is simulated and verified. If the simulation results show that the constraints are violated, the optimal adjustment strategy is locally readjusted through the pattern search algorithm to obtain the final optimal adjustment strategy, and the final optimal adjustment strategy is executed through the actuator.

[0188] In an optional embodiment, the deep learning model includes:

[0189] Construct an input layer to receive multi-dimensional real-time running data, connect a first convolution layer to the input layer, use a one-dimensional convolution operation in the first convolution layer, determine a small-scale convolution kernel according to a preset small-scale value, set the step size to a minimum value, keep the output and input sizes equal by padding, and apply a ReLU activation function to obtain a first convolution feature map, input the first convolution feature map into a first maximum pooling layer, set the pooling window size and step size of the first maximum pooling layer to the same value, and obtain a first pooling feature map;

[0190] Input the first pooled feature map into the first long short-term memory network layer, extract the time series feature according to the preset number of units to obtain the first long short-term memory feature, input the first long short-term memory feature into the second long short-term memory network layer, the number of units in the second long short-term memory network layer is half of the preset number of units, extract the long-term dependency feature to obtain the second long short-term memory feature, input the second long short-term memory feature into the global average pooling layer, map sequences of different lengths to vectors of preset fixed length, and obtain a global feature vector;

[0191] Input the global feature vector into the first fully connected layer, and obtain a first fully connected feature through a ReLU activation function; input the first fully connected feature into a Dropout layer, and obtain a random dropout feature based on a preset dropout rate; input the random dropout feature into a second fully connected layer, and set the number of neurons in the second fully connected layer to half the number of neurons in the first fully connected layer, and obtain a second fully connected feature through a ReLU activation function;

[0192] The second fully connected feature is input into the output layer, and the probability distribution of each abnormal category is obtained through the Softmax activation function.

[0193] First, multiple sensors are set at key locations of the system, including temperature sensors, pressure sensors, flow sensors, vibration sensors, etc. For example, temperature sensors and flow sensors are set on the steam extraction pipeline, pressure sensors and vibration sensors are set on the compressor and expander of the compressed air energy storage unit, and temperature sensors and flow sensors are set on the heat exchanger of the thermal energy storage unit. These sensors collect data at a preset sampling frequency (such as once per second).

[0194] After obtaining the real-time data from the sensor, the data is cleaned and preprocessed. Data cleaning includes removing outliers and missing values. For example, if the reading of a temperature sensor suddenly jumps from the normal 100°C to 1000°C, it can be determined as an outlier and removed. For missing values, the average value of the previous and next moments can be used to fill them. Preprocessing includes data normalization, which unifies data of different dimensions into the [0,1] interval for subsequent processing. After these processes, real-time operation data is obtained.

[0195] The real-time running data is fed into the deep learning model. The structure of the model is as follows:

[0196] The input layer receives multi-dimensional real-time operation data. Assuming there are 100 sensors and each sensor collects data for 10 time steps, the dimension of the input data is [100, 10].

[0197] The first convolutional layer is connected to the input layer and uses a one-dimensional convolution operation. Set a small-scale convolution kernel, such as kernel_size=3, stride=1, padding='same', to keep the output and input sizes equal. Use 32 convolution kernels and after the ReLU activation function, get the first convolution feature map with a dimension of [100,10,32].

[0198] The first maximum pooling layer processes the first convolutional feature map, sets pool_size = 2, stride = 2, and obtains the first pooled feature map, and the dimension becomes [50, 5, 32].

[0199] The first long short-term memory network layer receives the first pooled feature map, sets 64 LSTM units, extracts the time series features, and obtains the first long short-term memory features with a dimension of [50,64].

[0200] The second LSTM network layer processes the first LSTM features, sets 32 LSTM units (half of the first layer), extracts long-term dependency features, and obtains the second LSTM features with a dimension of [50,32].

[0201] The global average pooling layer maps the second longest short-term memory feature to a vector of fixed length to obtain a global feature vector with a dimension of [1,32].

[0202] The first fully connected layer receives the global feature vector, sets 64 neurons, uses the ReLU activation function, and obtains the first fully connected feature with a dimension of [1,64].

[0203] The Dropout layer processes the first fully connected feature and sets the dropout rate to 0.5 to obtain a random loss feature with the dimension unchanged, which is still [1,64].

[0204] The second fully connected layer receives the random dropout feature, sets 32 neurons (half of the first fully connected layer), uses the ReLU activation function, and obtains the second fully connected feature with a dimension of [1,32].

[0205] The output layer receives the second fully connected feature. Assuming that 5 abnormal categories are predefined, 5 neurons are set and the Softmax activation function is used to obtain the probability distribution of each abnormal category with a dimension of [1,5].

[0206] The deep learning model outputs the predicted status of each component of the system and determines the abnormal situation prediction results. For example, the model may output: the probability of normal operation is 0.8, the probability of a minor abnormality is 0.15, and the probability of a serious abnormality is 0.05.

[0207] Based on the abnormal situation prediction results, the cumulative deviations of multiple control parameters are calculated. For example, for the extraction steam flow control, the deviation between the actual flow and the target flow can be calculated and accumulated. Assume that in the past 10 minutes, the cumulative deviation has reached 50t / h.

[0208] When the cumulative deviation exceeds the preset control threshold, it is determined that the system operation state has undergone an over-threshold change. For example, if the preset extraction steam flow control threshold is 40 t / h, then when the cumulative deviation reaches 50 t / h, the online adaptive optimization mechanism is triggered.

[0209] After the online adaptive optimization mechanism is triggered, all control parameters are first reset to safe default values. For example, the steam extraction valve opening is adjusted to 50%, and the charge and discharge power of the compressed air energy storage unit and the flywheel energy storage unit is adjusted to 0.

[0210] At the same time, the sampling frequency of various sensors is increased to obtain high-frequency sampling data. For example, the original sampling frequency of once per second is increased to 10 times per second to obtain more detailed system operation status information.

[0211] Use recursive least squares to update the digital twin model based on high-frequency sampling data. The digital twin model is a virtual representation of the system, which contains the mathematical models of each component and their relationships. Through recursive least squares, the model parameters can be adjusted in real time to make it more accurately reflect the current system status.

[0212] Based on the updated current state of the digital twin model and external environment information (such as grid load, electricity price, heat load demand, etc.), the optimization objective function is redefined. For example, when the electricity price is high, more emphasis may be placed on increasing power generation; when the heat load demand is high, more emphasis may be placed on increasing the heating supply.

[0213] The optimal regulation strategy is re-determined and the strategy is simulated and verified. For example, the new regulation strategy may include: increasing the high-pressure steam extraction flow by 10%, reducing the medium-pressure steam extraction flow by 5%, discharging the compressed air energy storage unit at 30MW power to generate electricity, and discharging the flywheel energy storage unit at 10MW power.

[0214] If the simulation results show that a constraint is violated (for example, the temperature of a component exceeds a safe limit), the optimal regulation strategy is locally re-adjusted through a pattern search algorithm. The pattern search algorithm searches for a better solution near the current strategy until a strategy that satisfies all constraints is found.

[0215] Finally, the optimal regulation strategy that satisfies all constraints is obtained and executed through the actuators, which include steam extraction valves at each level, compressors and expanders of compressed air energy storage units, motors / generators of flywheel energy storage units, and heat exchanger control valves of thermal energy storage units.

[0216] In this embodiment, by setting up a variety of sensors, the system can monitor key parameters (such as temperature, pressure and flow) in real time, identify potential anomalies in time, and improve the reliability and safety of the system; data cleaning and preprocessing ensure the data quality of the input model, improve the accuracy of subsequent analysis, and thus enhance the predictive ability of the deep learning model; using the deep learning model to predict abnormal states makes the understanding of the states of each component of the system more accurate, and detects minor and serious anomalies in time, optimizing the response time; in abnormal situations, triggering the online adaptive optimization mechanism can quickly adjust the control parameters to prevent further deterioration of the system operation state and ensure system safety; increasing the sampling frequency and using the recursive least squares method to update the digital twin model make the virtual representation of the system more accurate, which can reflect the current state in real time and provide a reliable basis for decision-making; based on the updated model and external environment information, the optimization objective function is redefined, and the formulated adjustment strategy can be flexibly responded to in different environments, improving the overall efficiency; through the pattern search algorithm, it is ensured that the optimization adjustment strategy does not violate any safety constraints during execution, improving the feasibility and safety of the strategy; the flexible control of the actuator enables components such as steam extraction and energy storage units at all levels to work in coordination, optimize the overall system performance, and improve energy efficiency and economy.

[0217] Figure 2 FIG. 1 is a schematic diagram of the structure of an integrated regulation and control system for extraction steam energy storage and thermal energy storage according to an embodiment of the present invention. Figure 2 As shown, the system comprises:

[0218] The first unit is used to establish a digital twin model of an integrated system of steam extraction energy storage and thermal energy storage, wherein the digital twin model includes a steam turbine extraction system, an energy storage system and a thermal energy storage system; the steam turbine extraction system includes a multi-stage extraction point model and an extraction pipeline model; the energy storage system includes a compressed air energy storage unit model and a flywheel energy storage unit model; the thermal energy storage system includes a phase change material heat storage unit model and a molten salt heat storage unit model; and respectively collects a first real-time operating parameter corresponding to the steam turbine extraction system, a second real-time operating parameter corresponding to the energy storage system, a third real-time operating parameter corresponding to the thermal energy storage system, and a real-time load curve of the power grid, load forecast data for the next 24 hours, ambient temperature change trends, and renewable energy power generation forecast information;

[0219] The second unit is used to establish a multi-objective optimization model based on the digital twin model of the integrated system of extraction steam storage and thermal energy storage, the first real-time operating parameter, the second real-time operating parameter, the third real-time operating parameter, the real-time load curve of the power grid, the load forecast data for the next 24 hours, the ambient temperature change trend and the renewable energy power generation forecast information; using the improved multi-objective differential evolution algorithm, combined with the adaptive mutation strategy and the non-dominated sorting mechanism, the multi-objective optimization model is solved to obtain the optimal regulation strategy;

[0220] The third unit is used to send multi-stage steam extraction flow adjustment instructions to the steam turbine extraction system through the distributed control system based on the optimal adjustment strategy, and adopt the fuzzy PID control algorithm to adjust the steam extraction valve opening of each level of extraction point in real time to control the steam extraction flow; send coordinated charging and discharging instructions to the energy storage system, dynamically adjust the charging power and discharging power of the compressed air energy storage unit according to the load change of the power grid and the electricity price information, and control the charging and discharging power of the flywheel energy storage unit; send coordinated heat storage instructions to the thermal energy storage system, adaptively adjust the charging and discharging rates of the phase change material heat storage unit according to the heat load demand and the ambient temperature change, and control the charging and discharging flow of the molten salt heat storage unit.

[0221] According to a third aspect of the embodiments of the present invention,

[0222] An electronic device is provided, comprising:

[0223] processor;

[0224] a memory for storing processor-executable instructions;

[0225] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0226] A fourth aspect of the embodiments of the present invention is:

[0227] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0228] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0229] Finally, 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 aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. The integrated regulation and control method of steam extraction energy storage and thermal energy storage is characterized in that: include: Establish a digital twin model of an integrated system of steam extraction energy storage and thermal energy storage, wherein the digital twin model includes a steam turbine steam extraction system, an energy storage system and a thermal energy storage system; the steam turbine steam extraction system includes a multi-stage steam extraction point model and a steam extraction pipeline model; the energy storage system includes a compressed air energy storage unit model and a flywheel energy storage unit model; the thermal energy storage system includes a phase change material heat storage unit model and a molten salt heat storage unit model; respectively collect a first real-time operating parameter corresponding to the steam turbine steam extraction system, a second real-time operating parameter corresponding to the energy storage system, a third real-time operating parameter corresponding to the thermal energy storage system, as well as a real-time load curve of the power grid, load forecast data for the next 24 hours, ambient temperature change trends and renewable energy power generation forecast information; A multi-objective optimization model is established based on the digital twin model of the integrated system of extracted steam storage and thermal energy storage, the first real-time operating parameter, the second real-time operating parameter, the third real-time operating parameter, the real-time load curve of the power grid, the load forecast data for the next 24 hours, the ambient temperature change trend and the renewable energy power generation forecast information; an improved multi-objective differential evolution algorithm is used, combined with an adaptive mutation strategy and a non-dominated sorting mechanism, to solve the multi-objective optimization model and obtain the optimal regulation strategy; Based on the optimal regulation strategy, a multi-stage steam extraction flow regulation instruction is sent to the steam turbine extraction system through a distributed control system, and a fuzzy PID control algorithm is used to adjust the opening of the steam extraction valves at each level of extraction points in real time to control the steam extraction flow; a coordinated charging and discharging instruction is sent to the energy storage system, and the charging power and discharging power of the compressed air energy storage unit are dynamically adjusted according to the load changes of the power grid and the electricity price information, and the charging and discharging power of the flywheel energy storage unit is controlled at the same time; a coordinated heat storage instruction is sent to the thermal energy storage system, and the charging and discharging rates of the phase change material heat storage unit are adaptively adjusted according to the heat load demand and the ambient temperature changes, and the charging and discharging flow of the molten salt heat storage unit is controlled at the same time.

2. The method according to claim 1, characterized in that: A digital twin model of an integrated system of steam extraction energy storage and thermal energy storage is established, wherein the digital twin model includes a steam turbine steam extraction system, an energy storage system and a thermal energy storage system; the steam turbine steam extraction system includes a multi-stage steam extraction point model and a steam extraction pipeline model; the energy storage system includes a compressed air energy storage unit model and a flywheel energy storage unit model; the thermal energy storage system includes a phase change material heat storage unit model and a molten salt heat storage unit model, including: A digital twin model of a steam turbine extraction system is established, wherein the digital twin model includes a multi-stage extraction point model and an extraction pipeline model; the multi-stage extraction point model determines the relationship between the pressure and flow rate between the steam turbine stages based on the Stodola ellipse law, and calculates the steam parameters of each stage of extraction points in combination with the heat balance equation; the extraction pipeline model is based on the one-dimensional unsteady flow theory, combined with pipeline friction, heat loss and pressure fluctuation factors, and solves the steam flow state in the pipe through the characteristic line method; Establish a digital twin model of an energy storage system, wherein the digital twin model of the energy storage system includes a compressed air energy storage unit model and a flywheel energy storage unit model; the compressed air energy storage unit model is based on the isentropic compression-expansion theory, combined with the actual efficiency curves of the compressor and the turbine, simulates the energy conversion efficiency of the charging and discharging process, and determines the state change of the compressed air through the non-ideal gas state equation; the flywheel energy storage unit model is based on the law of conservation of angular momentum, combined with the dynamics model of the magnetic bearing system, to construct the acceleration process and deceleration process of the flywheel and determine the energy loss; A digital twin model of a thermal energy storage system is established, wherein the digital twin model of the thermal energy storage system includes a phase change material heat storage unit model and a molten salt heat storage unit model; the phase change material heat storage unit model adopts the enthalpy-porosity method, combined with the effective thermal conductivity, to simulate the heat transfer and phase change interface movement during the phase change process; the molten salt heat storage unit model is based on the computational fluid dynamics theory, adopts the three-dimensional non-steady-state heat conduction equation, and combines the influence of natural convection and thermal radiation to determine the molten salt temperature field distribution.

3. The method according to claim 1, characterized in that: The improved multi-objective differential evolution algorithm is used, combined with the adaptive mutation strategy and the non-dominated sorting mechanism, to solve the multi-objective optimization model and obtain the optimal adjustment strategy including: Set the population size and maximum number of iterations, randomly generate an initial feasible solution set that meets the constraints as the initial population, use the initial population as the current population, and start iteration: Performing non-dominated sorting on the current population, dividing the individuals in the current population into different non-dominated levels, calculating the crowding distance of individuals in the same non-dominated level in the current population, sorting the current population based on the non-dominated level and the crowding distance to obtain an ordered population, calculating the fitness improvement rate of the ordered population, and dynamically adjusting the mutation factor based on the fitness improvement rate; Use the adjusted mutation factor to perform mutation operation on each individual in the ordered population, randomly select three different individuals, generate mutant individuals through linear combination to obtain a mutant population, perform binomial crossover between the mutant population and the ordered population to generate a crossover population, and evaluate the objective function value of each individual in the crossover population based on the multi-objective optimization model; Merge the ordered population and the crossover population, perform non-dominated sorting and crowding distance calculation on the merged population again, obtain new non-dominated sorting results and new crowding distances, and select the best individuals from the merged population to form a new generation of population based on the new non-dominated sorting results and new crowding distances; The new generation population is taken as the current population, and the iteration is repeated until the preset maximum number of iterations is reached. The non-dominated solution set is extracted from the last generation population to determine the Pareto optimal solution set. The fuzzy comprehensive evaluation method is used to combine the decision maker's preferences and actual operating conditions to select the final optimal solution from the Pareto optimal solution set.

4. The method according to claim 1, characterized in that: Based on the optimal regulation strategy, a multi-stage steam extraction flow rate regulation instruction is sent to the steam turbine extraction system through a distributed control system, and the steam extraction valve opening of each level of steam extraction point is adjusted in real time by a fuzzy PID control algorithm to control the steam extraction flow rate; a coordinated charging and discharging instruction is sent to the energy storage system, and the charging power and discharging power of the compressed air energy storage unit are dynamically adjusted according to the load change of the power grid and the electricity price information, and the charging and discharging power of the flywheel energy storage unit is controlled at the same time; a coordinated heat storage instruction is sent to the thermal energy storage system, and the charging and discharging rate of the phase change material heat storage unit is adaptively adjusted according to the heat load demand and the ambient temperature change, and the charging and discharging flow rate of the molten salt heat storage unit is controlled at the same time, including: According to the multi-stage steam extraction flow dynamic regulation curve in the optimal regulation strategy, the actual steam extraction flow of each stage of steam extraction point is collected, the deviation and the deviation change rate from the target steam extraction flow are calculated, and the deviation and the deviation change rate are fuzzy processed to divide them into fuzzy subsets; Based on the fuzzy subset, an if-then fuzzy rule base is established, and the centroid method is used for defuzzification to obtain the adjustment amount of the proportional integral differential controller parameters, and the opening adjustment amount of the steam extraction valve is calculated by updating the proportional integral differential controller parameters, and the opening adjustment amount is sent to the corresponding steam extraction valve actuator through the distributed control system; Obtaining grid load data and electricity price information in real time, calculating the load deviation and electricity price gradient at the corresponding moment, and dividing the system operation state into multiple operation modes based on the corresponding values ​​of the load deviation and electricity price gradient, including peak-valley filling mode, peak shaving mode, valley filling mode and economic operation mode; Based on different operation modes, dynamically adjust the first charging and discharging power of the compressed air energy storage unit and the second charging and discharging power of the flywheel energy storage unit, and send the adjustment instructions of the first charging and discharging power and the second charging and discharging power to the corresponding charging and discharging actuators through the distributed control system; Collect heat load demand and ambient temperature in real time, calculate heat load deviation and ambient temperature change rate, and divide the operation state of the thermal energy storage system into a fast response mode, a stable operation mode and an energy-saving mode according to the corresponding values ​​of the heat load deviation and the ambient temperature change rate; The first current heat output of the phase change material heat storage unit and the second current heat storage of the molten salt heat storage unit are calculated; according to the ratio of the first current heat storage to the second current heat storage, the charging and discharging power distribution ratio of the phase change material heat storage unit and the molten salt heat storage unit are dynamically adjusted; based on the charging and discharging power distribution ratio, the charging and discharging rate of the phase change material heat storage unit and the charging and discharging flow rate of the molten salt heat storage unit are calculated, and sent to the corresponding charging and discharging actuators through the distributed control system.

5. The method according to claim 4, characterized in that The charging and discharging heat power distribution ratio is as follows: Among them, α represents the charging and discharging heat power allocation ratio, k1 represents the first weight, S PCM represents the current heat storage capacity of the phase change material heat storage unit, S salt represents the current heat storage capacity of the molten salt heat storage unit, k2 represents the second weight, S salt-max Indicates the maximum heat storage capacity of the molten salt heat storage unit; The heat charging and discharging rate of the phase change material heat storage unit is as follows: Among them, R PCM represents the charging and discharging rate of the phase change material heat storage unit, ΔQ represents the heat load deviation, and ΔH PCM Represents the latent heat of phase change of phase change material; The charging and discharging heat flow of the molten salt heat storage unit is as follows: Among them, F salt represents the heat charging and discharging flow rate of the molten salt heat storage unit, c p-salt represents the specific heat capacity of the molten salt, ΔT salt Represents the temperature difference of the molten salt.

6. The method according to claim 4, characterized in that The method further comprises: Setting up multiple sensors, acquiring real-time data from the sensors, performing data cleaning and preprocessing on the real-time data from the sensors to obtain real-time operation data, inputting the real-time operation data into a deep learning model, performing feature extraction on the real-time operation data, outputting the predicted status of each component of the system, and determining the abnormal situation prediction result; Based on the abnormal situation prediction results, the cumulative deviation of multiple control parameters is calculated. When the cumulative deviation exceeds the preset control threshold, it is determined that the system operation state has undergone an over-threshold change, triggering the online adaptive optimization mechanism, resetting all control parameters to safe default values, and increasing the sampling frequency of multiple sensors to obtain high-frequency sampling data; The digital twin model is updated based on the high-frequency sampling data using the recursive least squares method, and based on the current state of the model and external environment information, the optimization objective function is redefined, the optimal adjustment strategy is re-determined, and the optimal adjustment strategy is simulated and verified. If the simulation results show that the constraints are violated, the optimal adjustment strategy is locally readjusted through the pattern search algorithm to obtain the final optimal adjustment strategy, and the final optimal adjustment strategy is executed through the actuator.

7. The method according to claim 6, characterized in that The deep learning model includes: Construct an input layer to receive multi-dimensional real-time running data, connect a first convolution layer to the input layer, use a one-dimensional convolution operation in the first convolution layer, determine a small-scale convolution kernel according to a preset small-scale value, set the step size to a minimum value, keep the output and input sizes equal by padding, and apply a ReLU activation function to obtain a first convolution feature map, input the first convolution feature map into a first maximum pooling layer, set the pooling window size and step size of the first maximum pooling layer to the same value, and obtain a first pooling feature map; Input the first pooled feature map into the first long short-term memory network layer, extract the time series feature according to the preset number of units to obtain the first long short-term memory feature, input the first long short-term memory feature into the second long short-term memory network layer, the number of units in the second long short-term memory network layer is half of the preset number of units, extract the long-term dependency feature to obtain the second long short-term memory feature, input the second long short-term memory feature into the global average pooling layer, map sequences of different lengths to vectors of preset fixed length, and obtain a global feature vector; Input the global feature vector into the first fully connected layer, and obtain a first fully connected feature through a ReLU activation function; input the first fully connected feature into a Dropout layer, and obtain a random dropout feature based on a preset dropout rate; input the random dropout feature into a second fully connected layer, and set the number of neurons in the second fully connected layer to half the number of neurons in the first fully connected layer, and obtain a second fully connected feature through a ReLU activation function; The second fully connected feature is input into the output layer, and the probability distribution of each abnormal category is obtained through the Softmax activation function.

8. An integrated regulation and control system for extraction steam energy storage and thermal energy storage, used to implement the method described in any one of claims 1 to 7, characterized in that: include: The first unit is used to establish a digital twin model of an integrated system of steam extraction energy storage and thermal energy storage, wherein the digital twin model includes a steam turbine extraction system, an energy storage system and a thermal energy storage system; the steam turbine extraction system includes a multi-stage extraction point model and an extraction pipeline model; the energy storage system includes a compressed air energy storage unit model and a flywheel energy storage unit model; the thermal energy storage system includes a phase change material heat storage unit model and a molten salt heat storage unit model; and respectively collects a first real-time operating parameter corresponding to the steam turbine extraction system, a second real-time operating parameter corresponding to the energy storage system, a third real-time operating parameter corresponding to the thermal energy storage system, and a real-time load curve of the power grid, load forecast data for the next 24 hours, ambient temperature change trends, and renewable energy power generation forecast information; The second unit is used to establish a multi-objective optimization model based on the digital twin model of the integrated system of extraction steam storage and thermal energy storage, the first real-time operating parameter, the second real-time operating parameter, the third real-time operating parameter, the real-time load curve of the power grid, the load forecast data for the next 24 hours, the ambient temperature change trend and the renewable energy power generation forecast information; using the improved multi-objective differential evolution algorithm, combined with the adaptive mutation strategy and the non-dominated sorting mechanism, the multi-objective optimization model is solved to obtain the optimal regulation strategy; The third unit is used to send multi-stage steam extraction flow adjustment instructions to the steam turbine extraction system through the distributed control system based on the optimal adjustment strategy, and adopt the fuzzy PID control algorithm to adjust the steam extraction valve opening of each level of extraction point in real time to control the steam extraction flow; send coordinated charging and discharging instructions to the energy storage system, dynamically adjust the charging power and discharging power of the compressed air energy storage unit according to the load changes of the power grid and the electricity price information, and control the charging and discharging power of the flywheel energy storage unit; send coordinated heat storage instructions to the thermal energy storage system, adaptively adjust the charging and discharging rates of the phase change material heat storage unit according to the heat load demand and the ambient temperature changes, and control the charging and discharging flow of the molten salt heat storage unit.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method described in any one of claims 1 to 7.

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

Images