An efficiency optimization method and system for compressed air energy storage systems

By constructing a Kalman filter model and multidimensional nonlinear mapping equations to optimize the equipment speed and valve opening of the compressed air energy storage system, the problem of insufficient flow regulation and energy conversion efficiency of the compressed air energy storage system under complex power grid dispatching in the prior art is solved, and the system operation efficiency is optimized.

CN123116791APending Publication Date: 2026-11-17JINENG INT ENERGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611232149.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-14
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing compressed air energy storage systems struggle to achieve precise tracking of flow regulation and optimization of energy conversion efficiency in complex grid scheduling and dynamic changes in gas storage pressure, resulting in the system's operating efficiency failing to reach its optimal level.

Method used

By acquiring real-time gas storage pressure, compressor inlet temperature, and power grid dispatch commands, a Kalman filter model is constructed for data filtering, and isentropic efficiency and energy consumption per unit flow are calculated. Optimization is then performed using multidimensional nonlinear mapping equations, adjusting equipment speed and valve opening to optimize compression conditions.

Benefits of technology

It achieves a dual improvement in flow response accuracy and energy conversion efficiency under complex power grid dispatching, ensuring that the system operates with optimized efficiency within the limit of energy consumption increase, and meeting the needs of dynamic efficiency optimization and precise operating condition adjustment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123116791A_ABST
    Figure CN123116791A_ABST
Patent Text Reader

Abstract

This invention relates to the field of compressed air energy storage and power dispatching technology, and discloses an efficiency optimization method and system for compressed air energy storage systems. The method includes acquiring pressure change characteristics and performing noise reduction processing to obtain smooth state characteristics; calculating isentropic efficiency and unit flow energy consumption based on the smooth state characteristics and temperature parameters, and performing optimization processing accordingly to obtain initial equipment speed and initial valve opening; collecting real-time flow response data to calculate the time change gradient and obtain the adjustment deviation from the target flow gradient; if the deviation exceeds the deviation tolerance range, calculating compensation control commands under energy consumption constraints to correct and obtain the target equipment speed and target valve opening; adjusting the compression conditions according to the target parameters to obtain the actual efficiency optimization result. This method can achieve dynamic efficiency optimization and precise operating condition adjustment of compressed air energy storage systems, meeting the core requirements of improving both flow response accuracy and energy conversion efficiency.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the fields of compressed air energy storage and power dispatching technology, and in particular to an efficiency optimization method and system for compressed air energy storage systems. Background Technology

[0002] Currently, in the fields of compressed air energy storage and power dispatching technology, dynamic optimization of system operating efficiency, as a key link in reducing energy loss and improving response accuracy, has become a key research topic in the field.

[0003] In one existing technology, a closed-loop control method based on a distributed control system (DCS) or a supervisory control and data acquisition system (SCADA) combined with a frequency converter is used to achieve compressor operation regulation. Its control logic is based on conventional PID control feedback, for example, by acquiring gas storage pressure data through a data acquisition system and adjusting the compressor operation according to a preset static mapping curve or basic feedback loop. However, this conventional closed-loop control method is significantly inadequate in complex grid scheduling and dynamically changing gas storage pressure environments. Because conventional closed-loop parameters are mostly fixed, there is a lack of an online optimization mechanism that uses energy efficiency as the core objective variable. This easily leads to increased flow regulation deviation and decreased system energy conversion efficiency. Especially in complex scenarios with rapidly changing grid power commands and continuously fluctuating gas storage pressure, it is difficult to accurately track target flow changes and correct operating parameters within the optimal energy consumption range, resulting in significant lag in operation regulation and making it difficult for the overall system operating efficiency to reach the optimal level.

[0004] In summary, while existing technologies possess basic closed-loop control capabilities, they are still not intelligent enough to achieve dynamic efficiency optimization and precise operating condition adjustment of compressed air energy storage systems. They cannot meet the core requirements of monitoring and control and data control acquisition systems for improving both flow response accuracy and energy conversion efficiency in energy storage operation scenarios. Summary of the Invention

[0005] This invention provides an efficiency optimization method and system for compressed air energy storage systems, so as to realize dynamic efficiency optimization and precise operating condition adjustment of compressed air energy storage systems, and meet the core requirements of monitoring and control and data control acquisition systems for improving both flow response accuracy and energy conversion efficiency in energy storage operation scenarios.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides an efficiency optimization method for a compressed air energy storage system, comprising: The system acquires real-time gas storage pressure, compressor inlet temperature, thermal storage temperature, and grid dispatch commands; extracts the pressure change characteristics of the real-time gas storage pressure and performs noise reduction processing to obtain smooth state characteristics. Based on the smooth state characteristics, the compressor inlet temperature and the heat storage temperature, the isentropic efficiency and energy consumption per unit flow are calculated to obtain the real-time efficiency index. Then, optimization processing is performed based on the real-time efficiency index to obtain the initial equipment speed and the initial valve opening. The real-time flow response corresponding to the initial equipment rotation speed and initial valve opening is collected, the time change gradient is calculated to obtain the flow dynamic characteristics, and the adjustment deviation between the flow dynamic characteristics and the target flow gradient is calculated. If the adjustment deviation is greater than the deviation tolerance range, a compensation control command is calculated under the constraint that the estimated energy consumption after compensation does not exceed a preset multiple. The compensation control command is used to correct the initial equipment speed and initial valve opening to obtain the target equipment speed and target valve opening. If the deviation is not greater than the target speed and target valve opening, the initial equipment speed and initial valve opening are used as the target equipment speed and target valve opening. The compression conditions are adjusted based on the target equipment speed and target valve opening to obtain the actual efficiency optimization results.

[0007] In a second aspect, the present invention provides an efficiency optimization system for a compressed air energy storage system, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the efficiency optimization method for a compressed air energy storage system as described above.

[0008] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the efficiency optimization method for a compressed air energy storage system as described above.

[0009] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention obtains real-time gas storage pressure, compressor inlet temperature, thermal storage temperature and power grid dispatch instructions, extracts pressure fluctuation features through time matching and first-order difference, constructs a Kalman filter model to update time and measurement, and filters out high-frequency interference to obtain smooth state features. It integrates multi-source heterogeneous operating data and dynamic filtering algorithm to accurately restore the true thermodynamic state of the system, providing a high-confidence smooth state input for subsequent efficiency evaluation and optimization, and solving the problem of efficiency evaluation distortion caused by sensor noise and operating condition disturbances in the original operating data.

[0010] (2) This invention analyzes the actual compression ratio based on the smooth state characteristics, calculates the ideal isentropic compression work by combining temperature parameters and obtains the isentropic efficiency by combining the actual consumed shaft work, and calculates the energy consumption per unit flow to form a real-time efficiency index; extracts steady-state and transient components to establish a polynomial regression multidimensional nonlinear mapping equation, and obtains the initial equipment speed and valve opening by analytical optimization with the goal of minimizing energy consumption per unit flow under the boundary conditions of matching power grid dispatch instructions. It combines real-time efficiency quantitative evaluation with multivariate analytical optimization, and maximizes the thermodynamic efficiency of the compression process while meeting the power dispatch requirements, thus making up for the shortcomings of existing methods in that the control parameter setting lacks efficiency optimality consideration and the energy conversion efficiency is difficult to reach the theoretical optimal.

[0011] (3) This invention collects the real-time flow response corresponding to the initial parameters and calculates the time change gradient to obtain the adjustment deviation from the target flow gradient. When the deviation exceeds the tolerance range, under the constraint that the estimated energy consumption after compensation does not exceed a preset multiple, an identifier is generated based on the over-limit direction and response time difference, and the compensation control command is calculated to correct the target speed and opening degree. After safety and mechanical boundary verification, the actuator is synchronously driven and the energy consumption change rate is calculated to obtain the actual efficiency optimization result. While ensuring the accuracy of the flow response, the dynamic quality of the adjustment is also taken into account, which meets the core requirements of compressed air energy storage for improving both flow response accuracy and energy conversion efficiency under complex power grid scheduling. The energy consumption increase is strictly limited to the system operating efficiency within the operating efficiency boundary. Attached Figure Description

[0012] Figure 1 This is a schematic flowchart of an efficiency optimization method for a compressed air energy storage system provided in the first embodiment of the present invention; Figure 2 This is a schematic diagram of deviation adjustment provided in the first embodiment of the present invention; Figure 3 This is a schematic diagram of deviation exceeding limits and identification provided in the first embodiment of the present invention. Detailed Implementation

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

[0014] Reference Figure 1 The first embodiment of the present invention provides an efficiency optimization method for a compressed air energy storage system, comprising the following steps: S101, acquire real-time gas storage pressure, compressor inlet temperature, thermal storage temperature and grid dispatch instructions, extract the pressure change characteristics of the real-time gas storage pressure and perform noise reduction processing to obtain smooth state characteristics; S102, calculate the isentropic efficiency and unit flow energy consumption based on the smooth state characteristics, the compressor inlet temperature and the heat storage temperature to obtain the real-time efficiency index, and perform optimization processing based on the real-time efficiency index to obtain the initial equipment speed and initial valve opening. S103, collect the real-time flow response corresponding to the initial equipment rotation speed and initial valve opening, calculate the time change gradient to obtain the flow dynamic characteristics, and calculate the adjustment deviation between the flow dynamic characteristics and the target flow gradient. S104, if the adjustment deviation is greater than the deviation tolerance range, then a compensation control command is calculated under the constraint that the estimated energy consumption after compensation does not exceed a preset multiple, and the compensation control command is used to perform correction processing on the initial equipment speed and the initial valve opening to obtain the target equipment speed and the target valve opening; if it is not greater than, then the initial equipment speed and the initial valve opening are used as the target equipment speed and the target valve opening. S105, adjust the compression conditions according to the target equipment speed and target valve opening to obtain the actual efficiency optimization result.

[0015] In step S101, the acquisition of real-time gas storage pressure, compressor inlet temperature, thermal storage temperature, and grid dispatch commands, the extraction of pressure change characteristics of the real-time gas storage pressure, and the execution of noise reduction processing to obtain smooth state characteristics include: The system acquires real-time gas storage pressure, compressor inlet temperature, thermal storage temperature, and grid dispatch commands, and performs time-series matching processing to obtain a synchronized time series. The real-time gas storage pressure in the synchronous time series is calculated using first-order difference to obtain the pressure change sequence, and pressure fluctuation characteristics are extracted based on the pressure change sequence. The pressure fluctuation characteristics, the power grid dispatching instructions in the synchronous time series, the compressor inlet temperature and the thermal storage temperature execution characteristics are spliced ​​together to obtain the initial operating data. A Kalman filter model is constructed based on the initial running data, and the noise covariance matrix of the Kalman filter model is extracted. The initial running data is time-updated using the noise covariance matrix to obtain a priori state estimation sequence; a gain adjustment matrix is ​​calculated based on the priori state estimation sequence; and a measurement update process is performed on the priori state estimation sequence using the gain adjustment matrix to obtain a posterior state estimation sequence. High-frequency interference filtering is performed on the posterior state estimation sequence to obtain smooth state features.

[0016] It should be noted that, firstly, real-time gas storage pressure, compressor inlet temperature, and thermal storage temperature are acquired along with grid dispatch commands and time-series matching processing is performed to obtain a synchronized time series. Real-time gas storage pressure is acquired using a piezoresistive pressure sensor installed at a pre-reserved flange interface on the top of the gas storage tank. This sensor has a range of 0 to 15 MPa, an accuracy class of 0.1, and outputs a 4 to 20 mA standard current signal, which is converted into a digital quantity by an analog-to-digital converter module. The sampling frequency is set to 10 Hz. Compressor inlet temperature is acquired using a PT100 platinum resistance temperature sensor installed in the compressor intake pipeline. It has a range of 0 to 100℃, an accuracy class A (±0.15℃), and a sampling frequency of 1 Hz. Thermal storage temperature is acquired using a type K thermocouple installed at a temperature measuring point inside the thermal storage device. It has a range of 0 to 800℃, an accuracy class I (±1.5℃), and a sampling frequency of 1 Hz. The power grid dispatching instructions are received from the power grid dispatching system through the Ethernet communication interface. The communication protocol adopts the IEC 60870-5-104 protocol. The dispatching instructions are updated every minute. The content of the instructions is the active power target value that the charging compressor should perform, in MW, with a resolution of 0.1MW.

[0017] The timing matching process aligns data from four different sampling frequencies and update periods to a 1-second synchronization time base. Specifically, it performs arithmetic mean downsampling on the real-time gas storage pressure at 1-second intervals to merge 10 sampling points into a single per-second average value. It performs linear interpolation on the compressor inlet temperature and thermal storage temperature to upsample the 1Hz data to a 1-second interval. It performs zero-order hold on the grid dispatch command to copy and fill the updated value every minute into every second within that minute. This results in a synchronization time series with timestamps accurate to the second and corresponding values ​​for all four types of data at every second.

[0018] Next, first-order difference calculations are performed on the real-time gas storage pressure in the synchronous time series to obtain the pressure change sequence. Pressure fluctuation characteristics are then extracted from this sequence. The first-order difference calculation uses the difference between the pressure values ​​at two adjacent seconds in the synchronous time series as the output. The first-order difference at time t is the pressure value at time t minus the pressure value at time t-1, with the result in MPa / s. The pressure change sequence is composed of continuous first-order difference values ​​arranged chronologically, and its length is one sampling point shorter than the original pressure time series. The pressure fluctuation characteristics are measured using the root mean square (RMS) value of the difference sequence within a sliding window. The sliding window length is set to 60 sampling points, covering the difference data of the past 60 seconds. The RMS value is calculated by squaring all difference values ​​within the window, taking the arithmetic mean, and then taking the square root. This RMS value reflects the average fluctuation amplitude of the gas storage pressure change rate over the past 60 seconds, in MPa / s. The RMS value is calculated every second the window slides forward, and the output is a pressure fluctuation characteristic sequence updated synchronously with the synchronous time series.

[0019] Subsequently, the pressure fluctuation characteristics, grid dispatch instructions from the synchronous time series, compressor inlet temperature, and thermal storage temperature are subjected to feature concatenation processing to obtain the initial operating data. The concatenation process uses seconds as the time index, and at each whole second, the four feature values ​​are concatenated sequentially to form a four-dimensional feature vector. The four dimensions are, in order: pressure fluctuation characteristics (MPa / s), grid dispatch instructions (MW), compressor inlet temperature (°C), and thermal storage temperature (°C). The concatenated four-dimensional feature vector is then arranged in chronological order to form the initial operating data matrix. The number of rows in the matrix is ​​the length of the synchronous time series, and the matrix has four columns. For each column of the initial operating data matrix, min-max normalization is performed. During the normalization process, the normalization range of pressure fluctuation characteristics is set to 0 to 2 MPa / s and mapped to the interval 0 to 1. The range of power grid dispatch instructions is 0 to the upper limit of rated capacity and mapped to the interval 0 to 1. The range of compressor inlet temperature is -10 to 50℃ and mapped to the interval 0 to 1. The range of thermal storage temperature is 50 to 400℃ and mapped to the interval 0 to 1. The normalized matrix element values ​​all fall between 0 and 1.

[0020] Next, a Kalman filter model is constructed based on the initial operating data, and the noise covariance matrix of the Kalman filter model is extracted. The state variable of the Kalman filter model is the actual pressure value inside the gas storage tank, denoted as Ptrue, and the observation variable is the current pressure observation value after pressure fluctuation feature extraction, denoted as Pobs. The state transition equation approximates the time-varying law of the gas storage pressure as a first-order Markov process. The value of the state transition matrix A is determined by least-squares fitting of historical operating data. The fitting process uses approximately 4000 sets of continuous two-second pressure change data under similar operating conditions over the past 30 days to perform linear regression on the value of A. The process noise covariance matrix Q represents the deviation between the state transition model and the actual physical process, and the measurement noise covariance matrix R represents the deviation between the sensor measurement value and the actual pressure of the gas storage tank. The initial values ​​of Q and R are obtained by performing covariance statistical calculations on 2000 sets of sample data collected under calibration conditions. The calibration condition was set as follows: the gas storage pressure changed uniformly at a rate of 0.5 MPa / s within the range of 5 MPa to 10 MPa. Under this condition, the readings of a high-precision standard pressure gauge and an industrial pressure sensor were recorded simultaneously. The measurement noise sequence was obtained by subtracting the standard pressure gauge reading from the industrial sensor reading, and its variance was calculated as the initial value of R. The difference between two consecutive seconds of standard pressure gauge readings and the difference between A times the previous second reading was used as the process noise sequence, and its variance was calculated as the initial value of Q. Both Q and R matrices are 1-row, 1-column scalar positive values, which are kept fixed and not updated during subsequent online filtering to ensure the convergence and stability of the filter. For example, within the rated operating range of 5 MPa to 10 MPa gas storage pressure, linear regression was performed on approximately 4000 sets of consecutive two-second pressure change data over 30 days using the above least squares fitting method. The example value of the fitted state transition matrix A is approximately 0.997. This value is close to 1, indicating that the autocorrelation of the gas storage pressure within a 1-second interval is extremely high, which is consistent with the characteristics of a slow time-varying physical process of pressure in a gas storage system.

[0021] Then, time-update processing is performed on the initial operating data using the noise covariance matrix to obtain the prior state estimation sequence. The time-update processing is based on the prediction equation of the Kalman filter, recursively calculating second by second starting from the initial time. The prior state estimate for the first second is taken as the observed value at the current time as the starting point for filtering, and the prior error covariance is taken as the value of the measurement noise covariance R. From the second second onwards, the prior state estimate at the current time is equal to the state transition matrix A multiplied by the posterior state estimate from the previous time step, and the prior error covariance is equal to the square of A multiplied by the posterior error covariance from the previous time step plus the process noise covariance Q. Arranging the prior state estimates calculated for each second in chronological order constitutes the prior state estimation sequence, which represents the optimal prediction of the current pressure value based on historical states and the physical model.

[0022] Next, the gain adjustment matrix is ​​calculated based on the prior state estimation sequence. This gain adjustment matrix is ​​then used to perform measurement update processing on the prior state estimation sequence to obtain the posterior state estimation sequence. The formula for calculating the gain adjustment matrix K is: K equals the prior error covariance divided by the sum of the prior error covariance and the measurement noise covariance R. Since Q and R are both scalars, K is also a scalar value ranging from 0 to 1. The measurement update processing multiplies the gain K at that moment by the measurement residual (the difference between the observed value and the prior state estimate), and adds the product to the prior state estimate to obtain the posterior state estimate. The posterior error covariance is equal to 1 minus K, multiplied by the prior error covariance, and used for calculating the prior error covariance at the next moment. The posterior state estimation sequence output after performing the above measurement update every second is the optimal pressure state estimation result obtained by the Kalman filter after fusing the physical model prediction and sensor observations.

[0023] Subsequently, high-frequency interference filtering was performed on the posterior state estimation sequence to obtain smooth state features. High-frequency interference filtering employed a fourth-order Butterworth low-pass filter, with an infinite impulse response filter chosen to balance computational efficiency and filtering performance. The filter's cutoff frequency was set based on the frequency response characteristics of the gas storage system's mechanical actuators. The typical operating response frequencies of the gas tank inlet regulating valve and compressor are below 1Hz, while disturbances caused by pressure sensor noise and hydrodynamic transients are mainly concentrated in the frequency band above 2Hz. Therefore, high-frequency interference introduced by valve high-frequency oscillations and measurement noise was also attenuated. Since the Butterworth low-pass filtering was performed before the pressure data was downsampled to 1Hz (at which point the data sampling frequency was 10Hz and the Nyquist frequency was 5Hz), the cutoff frequency of the low-pass filter was preset to 1.5Hz to retain effective signal components reflecting the true trend of gas storage pressure changes.

[0024] The sampling frequency of the filter is set to 1Hz, consistent with the update frequency of the synchronous time series. The coefficients of the fourth-order Butterworth filter are calculated and determined using an analog filter design function with a cutoff frequency of 1.5Hz and a sampling frequency of 1Hz. The bilinear transform method is used to map the analog filter transfer function to the coefficients of the digital filter difference equation. The filter performs recursive calculations of the difference equation on the posterior state estimation sequence second by second, outputting a smoothed pressure state characteristic sequence. This characteristic sequence, together with the grid dispatch command, compressor inlet temperature, and thermal storage temperature in the synchronous time series, constitutes the smoothed state characteristics.

[0025] In step S102, the calculation of isentropic efficiency and energy consumption per unit flow rate based on the smooth state characteristics, the compressor inlet temperature, and the heat storage temperature to obtain real-time efficiency indicators includes: The current actual compression ratio is analyzed based on the smooth state characteristics; The ideal isentropic compression work is calculated by combining the compressor inlet temperature, the heat storage temperature, and the actual compression ratio. The actual shaft work consumed by the compressor is obtained, and the ratio of the ideal isentropic compression work to the actual shaft work consumed is calculated to obtain the isentropic efficiency. The actual gas mass flow rate is obtained, and the ratio of the actual shaft work consumed to the actual gas mass flow rate is calculated to obtain the energy consumption per unit flow rate. The isentropic efficiency and the energy consumption per unit flow are used as real-time efficiency indicators.

[0026] It should be noted that the current actual compression ratio is first analyzed based on the smoothed state characteristics. The smoothed state characteristics are output in step S101, and include a gas storage pressure state estimation sequence after dual processing with Kalman filtering and low-pass filtering. The actual compression ratio is defined as the ratio of the compressor discharge pressure to the intake pressure. The discharge pressure is taken as the gas storage pressure value at the current moment in the smoothed state characteristics, in MPa. The intake pressure is directly measured by a pressure sensor installed in the compressor intake pipe. The sensor has a range of 0 to 1 MPa and an accuracy class of 0.1. This pressure value has been included in the synchronous time series through time-series matching in step S101. The current actual compression ratio is calculated by dividing the current discharge pressure value by the intake pressure value at the same moment in the synchronous time series. The result is a dimensionless value greater than 1, rounded to three decimal places.

[0027] When the compressor is stopped or the intake valve is not open, the suction pressure approaches atmospheric pressure, which is about 0.101 MPa. At this time, the physical meaning of the actual compression ratio is the multiple of the storage pressure relative to atmospheric pressure.

[0028] Next, the ideal isentropic compression work is calculated by combining the compressor inlet temperature, the heat storage temperature, and the actual compression ratio. The compressor inlet temperature is extracted from the synchronous time series in step S101, and its value is the measurement value of the PT100 temperature sensor at the compressor inlet at the current moment, in °C. Before calculation, it needs to be converted to thermodynamic temperature in K, by adding 273.15 to the °C value. The heat storage temperature is also extracted from the synchronous time series, and its value is the measurement value of the K-type thermocouple inside the heat storage device at the current moment. It is used to determine the thermal influence of the heat storage system on the compression process in subsequent steps. However, in the calculation of the ideal isentropic compression work, the heat storage temperature is not directly involved in the formula calculation but is reserved as input for the thermodynamic state assessment in step S103.

[0029] The calculation of ideal isentropic compression work is based on the assumption that air is an ideal gas and undergoes an isentropic process. The key parameter in the calculation formula is the specific heat ratio of air, denoted as . The value is taken as 1.4, and the gas constant is denoted as... The value is 0.287 kJ / (kg·K). Ideal isentropic compression work. The calculation formula is as follows: ; In the formula: The work done by ideal isentropic compression is expressed in kJ / kg. It is the gas constant; The temperature at the compressor inlet is the converted thermodynamic temperature, expressed in Kelvin (K). The specific heat capacity ratio of air; This represents the actual compression ratio. The unit of the calculated result is kJ per kg, which represents the mechanical work required to compress one kilogram of air under ideal, lossless, isentropic conditions, rounded to two decimal places. The dimensions on both sides of the equation in this formula are consistent: the left side is kJ per kg, and the right side... The unit is kJ per kg per K. Multiplying this by the temperature unit K and then by the dimensionless specific heat capacity ratio function also yields kJ per kg.

[0030] Subsequently, the actual shaft work consumed by the compressor is acquired, and the ratio of the ideal isentropic compression work to the actual shaft work consumed is calculated to obtain the isentropic efficiency. The actual shaft work consumed is measured by a power transmitter installed at the input of the compressor drive motor. The power transmitter simultaneously collects the instantaneous values ​​of three-phase voltage and three-phase current and calculates the active power. The sampling frequency is 1Hz, aligned with the synchronous time series, and the output unit is kW. To ensure consistency with the unit of ideal isentropic compression work, the active power output by the power transmitter is divided by the actual gas mass flow rate passing through the compressor at the current moment to obtain the actual shaft work consumed per unit mass of air, also in kJ per kg. The isentropic efficiency is calculated by dividing the ideal isentropic compression work by the aforementioned actual shaft work consumed per unit mass; the result is a dimensionless ratio between 0 and 1, rounded to four decimal places. A closer isentropic efficiency to 1 indicates a smaller deviation of the compressor from the ideal isentropic process during actual operation, resulting in lower internal friction losses, flow separation losses, and leakage losses. If the actual shaft work consumed per unit mass is 220.30 kJ per kg, then the isentropic efficiency is 0.8193, approximately 81.93%. For example, under operating conditions with a compressor inlet temperature of 300 K (approximately 26.85 °C) and an actual compression ratio of 5.2, the calculated ideal isentropic compression work is 180.50 kJ per kg. Then, the actual gas mass flow rate is obtained, and the ratio of the actual shaft work consumed to the actual gas mass flow rate is calculated to obtain the energy consumption per unit flow rate. The actual gas mass flow rate is directly measured by a thermal mass flow meter installed on the compressor outlet pipeline. The flow rate measurement range is 0 to 20 kg / s, the accuracy class is 0.5, and the output signal is a 4 to 20 mA analog signal updated at a frequency of 1 Hz after analog-to-digital conversion. The energy consumption per unit flow rate is calculated by dividing the actual shaft work consumed (in kW) output by the power transmitter by the current reading of the actual gas mass flow rate (in kg / s), with the result in kJ per kg, rounded to two decimal places. The physical meaning of energy consumption per unit flow rate is the actual electrical power required to compress one kilogram of air. The lower this value, the less electrical energy is required to compress the same mass of air, and the higher the real-time operating efficiency of the system.

[0031] Finally, the isentropic efficiency and energy consumption per unit flow are assembled and output as real-time efficiency indicators. Every second, a set of isentropic efficiency and energy consumption per unit flow values ​​for the current moment are calculated. The two values ​​are packaged into a two-element structure in floating-point format. The structure definition contains three fields: timestamp, isentropic efficiency, and energy consumption per unit flow. The timestamp is accurate to the second to maintain alignment with the synchronized time series.

[0032] In one implementation, an optimization process is performed based on the real-time efficiency index to obtain the initial equipment speed and the initial valve opening, including: Extract the steady-state and transient components from the smooth state features; The steady-state component, the transient component, and the real-time efficiency index are input into a preset polynomial regression fitting model to perform fitting calculations and establish a multidimensional nonlinear mapping equation between efficiency and flow output. Based on the multidimensional nonlinear mapping equation, under the boundary condition of ensuring that the flow output matches the power grid dispatch command, analytical optimization is performed with the goal of minimizing the energy consumption per unit flow of the system to obtain the initial equipment speed and the initial valve opening.

[0033] Subsequently, under the boundary condition of ensuring that the flow output matches the grid dispatch command, analytical optimization is performed with the goal of minimizing the energy consumption per unit flow of the system. Analytical optimization uses a multidimensional nonlinear mapping equation as the objective function, substitutes the target flow value corresponding to the grid dispatch command into the flow balance equation as an equality constraint, and uses the continuous operating range of equipment speed (from minimum continuous speed to 105% of rated speed) and the full stroke range of valve opening (0% to 100%) as upper and lower bound constraints for variables. The Lagrange multiplier method is used to transform the constrained optimization problem into an unconstrained optimization problem for solution, resulting in a combination of equipment speed and valve opening that satisfies all constraints and minimizes the energy consumption per unit flow. This combination represents the initial equipment speed and initial valve opening. ; In the formula, and These are the initial equipment speed and initial valve opening obtained through optimization, respectively; It is the bias vector; The coefficient matrix is ​​formed by fitting historical operating data using the least squares method. For steady-state components, For transient components, For isentropic efficiency, The energy consumption per unit flow rate forms a four-dimensional input feature vector. The fitting data is selected from continuous operating records of similar compressor units under different operating conditions over the past 30 days, with approximately 4000 data points. Cross-validation was used to select the polynomial order during fitting to avoid overfitting. The resulting multidimensional nonlinear mapping equation establishes a nonlinear functional relationship between the input variables and the equipment speed and valve opening.

[0034] Next, the steady-state component, transient component, and isentropic efficiency and energy consumption per unit flow rate from the real-time efficiency index are used as input features and input into a pre-defined second-order polynomial regression fitting model. The input to the polynomial regression fitting model is a four-dimensional feature vector, and the output is a two-dimensional vector composed of equipment speed and valve opening. The model form is as follows: It should be noted that, firstly, the steady-state and transient components are extracted from the smoothed state features. The steady-state component is obtained by performing a moving average filter on the smoothed state feature sequence, with a moving window length of 30 sampling points, corresponding to a 30-second data span. This window length is selected based on the pressure response time constant of the gas storage system. The transient component is obtained by subtracting the corresponding steady-state component value from the values ​​at each time step of the smoothed state feature sequence; both have the same dimensions as the smoothed state features.

[0035] For example, under the nominal operating conditions (gas storage pressure approximately 7.5 MPa, compressor inlet temperature approximately 300 K, and heat storage temperature approximately 350 K) of a geared centrifugal compressor unit with a rated power of 50 MW, the bias vector obtained by the above-mentioned second-order polynomial regression fitting is... Example values ​​are approximately In the coefficient matrix W, the example value of the coefficient corresponding to the steady-state component to the rotational speed is approximately 1250, and the example value of the coefficient corresponding to the isentropic efficiency to the rotational speed is approximately 580. Under this operating condition, after substituting the normalized value of the steady-state component (0.52), the normalized value of the transient component (0.03), the isentropic efficiency (0.82), and the energy consumption per unit flow rate (185.2 kJ / kg) into the fitting equation, the predicted initial equipment rotational speed is approximately 2980 r / min, and the initial valve opening is approximately 62%.

[0036] In step S103, the real-time flow response corresponding to the initial equipment rotation speed and initial valve opening is acquired, the time change gradient is calculated to obtain the flow dynamic characteristics, and the adjustment deviation between the flow dynamic characteristics and the target flow gradient is calculated, including: Based on the initial equipment rotation speed and the initial valve opening, a drive command is generated and sent to the underlying controller, and the real-time flow response of the compressor exhaust port is collected; The real-time traffic response is subjected to data cleaning processing to obtain an effective traffic sequence; Extract adjacent time nodes from the effective traffic sequence; Calculate the ratio of the flow difference between adjacent time nodes to the time interval to obtain the time change gradient; Perform feature aggregation processing on the time-varying gradient to obtain dynamic traffic features; The difference between the dynamic characteristics of the flow and the target flow change gradient corresponding to the power grid dispatch command is calculated to obtain the adjustment deviation.

[0037] It should be noted that, firstly, a drive command is generated based on the initial equipment speed and initial valve opening and sent to the underlying controller, while the real-time flow response of the compressor exhaust port is collected. The initial equipment speed and initial valve opening are output by the optimization process in step S102. The speed unit is r / min, and the opening unit is the percentage range of 0% to 100%, corresponding to the valve being fully closed to fully open. When the drive command is generated, the speed value is written to the frequency setting register of the frequency converter via the Modbus RTU protocol. The communication baud rate is 9600bps, and the data format is 8 data bits, 1 stop bit, and no parity. The frequency converter controls the operating frequency of the drive motor according to the speed setting value. The valve opening command is sent to the electric valve positioner through a 4 to 20mA analog output module. 4mA corresponds to 0% opening, and 20mA corresponds to 100% opening. The positioner's built-in PID controller drives the servo motor to adjust the valve core position to the target opening. The real-time flow response of the compressor exhaust port is acquired by an ultrasonic flow meter installed on the exhaust pipe. The ultrasonic flow meter adopts the time difference method measurement principle, with a range of 0 to 50 m³ / s, an accuracy class of 1.0, and a sampling frequency of 10 Hz. The output signal is transmitted to the upper data acquisition unit via RS485 interface using the Modbus ASCII protocol. Each message contains the instantaneous flow value and a timestamp. The instantaneous flow value is in m³ / s and is rounded to two decimal places.

[0038] Next, data cleaning is performed on the real-time flow response to obtain the effective flow sequence. Ultrasonic flow meters may output abnormal jumps beyond the physically reasonable range during compressor start-up and shutdown or when there is severe airflow disturbance in the pipeline. The first step of data cleaning is range verification, marking data points with instantaneous flow values ​​less than 0 m³ / s or greater than the upper limit of the flow meter's range (50 m³ / s) as invalid and directly discarding them. The second step of data cleaning is volatility verification, calculating the absolute value of the flow change between the current sampling point and the previous valid sampling point, divided by the time interval to obtain the absolute value of the rate of change per unit time. If this absolute value exceeds 50 m³ / s², the current sampling point is judged as an abnormal abrupt change and discarded. The 50 m³ / s² limit is based on the physical constraint of the maximum rate of change of the compressor's exhaust flow under extreme operating conditions. The measured maximum rate of change of the exhaust flow when the compressor abruptly changes from rated speed to maximum speed does not exceed 30 m³ / s², and this limit is obtained by taking a safety margin of 1.7. The third step in data cleaning is smoothness verification. For data points that pass the verification in the first two steps, a moving median filter with a window size of 5 sampling points is applied. The median of the flow rate values ​​within the window is used as the replacement value for the center position of the window to suppress random high-frequency electronic noise. The filtered sequence is the effective flow rate sequence.

[0039] Subsequently, adjacent time nodes of the effective flow sequence are extracted. The timestamps of each sampling point in the effective flow sequence are marked by the ultrasonic flow meter at the time of acquisition, with an accuracy of 0.1 seconds. The time interval between two adjacent sampling points is determined by the sampling frequency of 10Hz, i.e., fixed at 0.1 seconds. Adjacent time nodes are extracted by sequentially reading the timestamp of each element starting from the second element of the effective flow sequence and the timestamp of the previous element, forming time node pairs for subsequent gradient calculation. Simultaneously, the corresponding two flow values ​​are read to form flow value pairs.

[0040] Then, the ratio of the flow rate difference between adjacent time nodes to the time interval is calculated to obtain the time variation gradient. For each pair of adjacent time nodes, the flow rate difference is obtained by subtracting the flow rate of the previous time node from the flow rate of the later time node, with the time interval fixed at 0.1 seconds. The flow rate variation gradient within that time interval is obtained by dividing the two values, with the unit being m³ / s². After traversing all adjacent node pairs in the effective flow rate sequence and calculating them one by one, a set of time variation gradient sequences with the same length as the effective flow rate sequence minus one is obtained. This sequence reflects the instantaneous rate of change of the compressor exhaust flow rate within each 0.1-second interval.

[0041] Next, feature aggregation processing is performed on the time-varying gradients to obtain the flow dynamics characteristics. The purpose of feature aggregation is to condense the discrete instantaneous gradient sequences into feature indicators that can comprehensively characterize the current flow dynamic response capability. The aggregation method is as follows: the time-varying gradient sequences corresponding to 20 sampling points within the past 2 seconds are taken as the current analysis window. The root mean square (RMS) of all gradient values ​​within this window is calculated. The RMS calculation process involves squaring each gradient value within the window, taking the arithmetic mean, and then taking the square root. The RMS gradient value can reflect the effective average rate of flow change within the window period, while also suppressing individual spikes. In addition, the maximum and minimum values ​​of gradient values ​​within the window are calculated simultaneously. The three statistics—RMS gradient, maximum gradient, and minimum gradient—are concatenated into a three-dimensional feature vector, which is the flow dynamics characteristic at the current moment. The flow dynamics characteristic is updated every 0.1 seconds, when the window slides forward by one sampling point.

[0042] The difference between the dynamic flow characteristics and the target flow change gradient corresponding to the grid dispatch command is then calculated to obtain the adjustment deviation. The target flow change gradient corresponding to the grid dispatch command is obtained by parsing the grid dispatch command acquired in step S101. The grid dispatch command contains a sequence of active power target values ​​that the compressor should execute. The difference between the active power target value at the current moment and the previous moment, divided by the time interval, yields the power change rate, in MW / min. The conversion between the power change rate and the flow change gradient is achieved through the power-flow characteristic curve under the compressor's rated operating conditions. This characteristic curve records the shaft work required by the compressor to output a unit mass flow rate at various speed and pressure combinations. The linear proportionality coefficient between the power and the exhaust flow rate at rated pressure is taken as the conversion coefficient. The power change rate is divided by this conversion coefficient and then by the gas density constant to obtain the target flow change gradient, also in m³ / s². The difference between the root mean square gradient in the current flow dynamic characteristics and the target flow change gradient is the adjustment deviation, in m³ / s², rounded to two decimal places.

[0043] like Figure 2 As shown in the figure, this figure illustrates the time-series distribution of the adjustment deviation of the compressor exhaust flow dynamic response after being driven by the initial equipment speed and initial valve opening. The figure clearly shows that the deviation (blue solid line) between the actual flow gradient and the target flow gradient before compensation correction frequently jumps within a large range of ±2 m³ / s², and repeatedly exceeds the deviation tolerance range of ±0.05 m³ / s² (upper and lower limits of the red dashed line) at approximately 40, 80, 120, and 160 seconds, with peak deviations exceeding ±2 m³ / s². This result intuitively reveals that the initial parameters obtained solely through optimization in step S102 still exhibit significant flow tracking lag and overshoot risks under complex operating conditions, providing a solid foundation of deviation data for the calculation and correction of the compensation control command in subsequent step S104.

[0044] In step S104, if the adjustment deviation is greater than the deviation tolerance range, a compensation control command is calculated under the constraint that the estimated energy consumption after compensation does not exceed a preset multiple, including: The adjustment deviation is subjected to feature decomposition processing to obtain the over-limit direction feature and response time difference; If the over-limit direction feature is positive and the response time difference is less than a preset time threshold, a flow overshoot flag is generated; if the over-limit direction feature is negative or the response time difference is not less than the preset time threshold, a response lag flag is generated. Based on the flow overshoot flag or the response hysteresis flag, the adjustment deviation is multiplied by the current pressure to obtain the basic compensation base. Based on the flow overshoot flag or the response hysteresis flag, the corresponding weights are assigned to the basic compensation base to obtain the primary speed compensation amount and the primary valve compensation amount. Energy consumption simulation calculations are performed using the primary speed compensation amount and the primary valve compensation amount to obtain the estimated energy consumption per unit flow rate. If the estimated unit flow energy consumption does not exceed the preset multiple before adjustment, then the primary speed compensation amount and the primary valve compensation amount are used as the speed compensation amount and valve compensation amount; if they exceed, then the primary speed compensation amount and the primary valve compensation amount are subjected to proportional attenuation processing to obtain the speed compensation amount and valve compensation amount that meet the constraint conditions. The speed compensation amount and the valve compensation amount are spliced ​​together to obtain a compensation control command.

[0045] It should be noted that, firstly, feature decomposition is performed on the regulation deviation to obtain the out-of-limit directional feature and the response time difference. The regulation deviation is calculated and output from the previous steps, with units of m³ / s². A positive value indicates that the actual flow rate change gradient is less than the target gradient, i.e., insufficient response; a negative value indicates that the actual flow rate change gradient is greater than the target gradient, i.e., over-response. The out-of-limit directional feature is extracted by determining the sign of the regulation deviation at the current moment. If it is positive, the out-of-limit directional feature is assigned a value of 1, indicating positive deviation; if it is negative, the out-of-limit directional feature is assigned a value of -1, indicating negative deviation; and if it is zero, it is assigned a value of 0, indicating that it is exactly following the target. The response time difference is extracted by starting the timer from the moment the adjustment deviation first exceeds the tolerance range. The tolerance range is set based on the engineering control precision of the compressor flow regulation. The steady-state error band of the compressor's flow gradient regulation under rated operating conditions is set to ±0.05 m³ / s². This error band is jointly determined by the minimum resolution of the compressor inverter speed closed-loop control and the repeatability of the valve positioner. An inverter resolution of 0.1 r / min corresponds to a flow gradient of approximately 0.03 m³ / s², and a valve positioner repeatability of 0.5% corresponds to a flow gradient of approximately 0.02 m³ / s². The two are combined using the root mean square and rounded to obtain 0.05 m³ / s². The number of seconds elapsed from the moment the deviation exceeds the tolerance range to the current moment is the response time difference, accurate to 0.1 seconds. It is obtained by directly accumulating the sampling interval count from the flow dynamic characteristic timestamp sequence output in step S103.

[0046] Next, an indicator generation and judgment are performed. If the over-limit direction characteristic is positive and the response time difference is less than the preset time threshold, a flow overshoot indicator is generated. The preset time threshold is based on the settling time of the compressor exhaust flow under a typical step response. After applying a 10% speed step change to the compressor, the transition time required for the exhaust flow to reach 90% of its steady-state value from the initial value is statistically estimated to be approximately 1.5 to 2.5 seconds. The median value of 2 seconds is taken as the preset time threshold. If the response time difference is less than 2 seconds, the normal transition process has not yet been completed, and the flow is still in the climbing stage, but a positive deviation has already occurred. This indicates that the control overshoot has caused the flow to exceed the target and continue to rise, which can be judged as a flow overshoot state. If the over-limit direction characteristic is negative or the response time difference is not less than the preset time threshold, a response lag indicator is generated. A negative deviation itself indicates that the actual gradient lags behind the target, which is a response lag. A positive deviation, but with a response time difference exceeding 2 seconds, indicates that the flow has not reached the target after the normal transition time, which is also a response lag. This preset time threshold can be adjusted within the range of 1.5 to 3 seconds according to the specific compressor model and pipeline length.

[0047] Subsequently, based on the flow overshoot flag or response lag flag, the regulation deviation is multiplied by the current pressure to obtain the basic compensation base. The current pressure is taken as the gas storage pressure value at the current moment from the smoothing state characteristics output in step S101, in MPa. The basic compensation base is calculated by multiplying the absolute value of the regulation deviation by the current pressure value, in MPa·m³ / s², and then adding a directional correction according to the flag type. Under the flow overshoot flag, the basic compensation base is negative to reduce overshoot in subsequent outputs; under the response lag flag, the basic compensation base is positive to improve compensation lag in subsequent outputs. This calculation method ensures that the compensation amount simultaneously considers the magnitude of the flow tracking error and the current system pressure level. The higher the pressure, the larger the compensation base for the same flow deviation, so as to provide stronger regulation in the high-pressure section.

[0048] Then, based on the flow overshoot or response lag indicator, corresponding weights are assigned to the basic compensation base to obtain the primary speed compensation and primary valve compensation. The weight allocation is calculated based on the compressor's power angle characteristics, determined by the ratio of the sensitivity of the gear centrifugal compressor's speed increase to the flow change under a certain power condition to the sensitivity of its opening increase to the flow change. In historical test data near the rated operating condition, flow overshoot requires a rapid reduction in output, prioritizing valve retraction to quickly reduce flow while supplementing with speed correction. Therefore, under the flow overshoot indicator, the valve weight is set to 0.65, and the speed weight is set to 0.35. Response lag requires an increase in output, but prioritizes speed increase to improve the compression ratio. Therefore, under the response lag indicator, the valve weight is set to 0.55, and the speed weight is set to 0.45. The primary speed compensation is calculated by multiplying the basic compensation base by the speed weight, and the primary valve compensation is calculated by multiplying the basic compensation base by the valve weight. The unit of speed compensation is r / min, and the unit of valve compensation is opening percentage. A 1% increase in rotational speed results in a flow rate change of approximately 0.7%, while a 1% increase in valve opening results in a flow rate change of approximately 1.0%. The sensitivity ratio between the two is approximately 1.0 / 0.7 ≈ 1.43. The aforementioned weights are determined by using a sensitivity ratio of 1.43 as a baseline, and adjusting the baseline ratio according to the control strategy preferences under two scenarios: flow overshoot and response lag. In the flow overshoot scenario, it is necessary to prioritize reducing the flow output, and the valve action is more sensitive to the impact on flow. Therefore, the valve weight ratio is increased to 0.65 / 0.35≈1.86 based on the baseline ratio. In the response lag scenario, it is necessary to consider increasing the rotational speed to improve the compression ratio. Therefore, the valve weight ratio is appropriately reduced to 0.55 / 0.45≈1.22.

[0049] It should be noted that when converting the basic compensation base (unit: MPa·m³ / s²) to speed compensation (unit: r / min) and valve compensation (unit: percentage of opening), dimensional uniformity is achieved through pre-calibrated scale conversion coefficients. The speed conversion coefficient is determined by the slope of the compressor's speed-power characteristic curve under rated pressure conditions, and the valve conversion coefficient is determined by combining the input current-opening mapping relationship of the electric valve positioner with the compressor's flow-opening characteristics. These scale conversion coefficients are implicitly incorporated into the corresponding weight values ​​during the weighting calculation; therefore, multiplying the basic compensation base by the weights directly yields the correct compensation amount in physical units.

[0050] Next, energy consumption simulation calculations are performed using primary speed compensation and primary valve compensation to obtain the estimated energy consumption per unit flow rate. The energy consumption simulation is based on the compressor's steady-state performance regression formula, which is obtained by fitting compressor factory performance test data and on-site debugging records, recording the isentropic efficiency and shaft power consumption under different combinations of speed and valve opening. The current speed plus the primary speed compensation and the current valve opening plus the primary valve compensation are used as the simulation input conditions. Four discrete operating point values ​​closest to the speed and valve opening are found in the regression formula. The estimated shaft power and estimated isentropic efficiency under this simulation condition are calculated using bilinear interpolation. The interpolation order is to perform two linear interpolations on the speed dimension and one linear interpolation on the valve opening dimension. The estimated shaft power is divided by the actual gas mass flow rate at the current moment, i.e., the instantaneous mass flow rate value measured by the ultrasonic flow meter in step S102, to obtain the estimated energy consumption per unit flow rate, in kJ per kg, rounded to two decimal places.

[0051] Subsequently, energy consumption constraint judgment is performed. The preset multiplier, based on the compressed air energy storage power station's settings, multiplies the unit flow energy consumption before adjustment by a maximum allowable increase multiplier to determine the upper limit of energy consumption. The maximum allowable increase multiplier is 1.2, calculated by the power station. When the unit energy consumption increase exceeds 20%, the electro-to-electro-electric conversion efficiency of the energy storage cycle will drop below 55%. If the estimated unit flow energy consumption does not exceed 1.2 times the unit flow energy consumption before adjustment, the primary speed compensation and primary valve compensation are directly output as the speed compensation and valve compensation. If the estimated energy consumption per unit flow exceeds 1.2 times the upper limit, a proportional attenuation process is applied to the primary speed compensation and primary valve compensation. The attenuation method involves multiplying both the primary speed compensation and primary valve compensation by an attenuation coefficient. This coefficient is calculated by multiplying the energy consumption per unit flow before adjustment by 1.2 and then dividing by the estimated energy consumption per unit flow; the result is a positive value less than 1. After attenuation, the two compensation amounts are reduced proportionally to re-meet the energy consumption constraints. The results after attenuation calculation are the speed compensation and valve compensation amounts that satisfy the constraints. (Operating efficiency red line: Power plant energy efficiency constraint model, safe operation bottom line) Finally, the speed compensation and valve compensation are concatenated to obtain the compensation control command. The concatenation method involves writing the speed compensation into the high 8 bits of a 16-bit integer variable and the valve compensation into the low 8 bits of the same variable. Before writing, the valve compensation must be mapped from the percentage range to an integer value between 0 and 255. This 16-bit integer variable is then assembled sequentially with the command header and check frame. The command header, occupying one byte and set to 0xAA, identifies the type of compensation control command. The check frame, also occupying one byte, is the result of the XOR operation between the command header and the data variable, used for communication verification. The complete command frame is 4 bytes long and is sent to the underlying controller for execution via the Modbus RTU protocol. After parsing the command, the underlying controller adds the speed compensation to the current speed setpoint and adds the valve compensation to the corresponding register of the current valve opening setpoint, completing one closed-loop control cycle of compensation adjustment.

[0052] In one implementation, the step of using the compensation control command to perform correction processing on the initial equipment speed and initial valve opening to obtain the target equipment speed and target valve opening includes: The compensation control command is parsed to extract the speed correction range and opening correction range; The initial target speed is obtained by superimposing the speed correction range with the initial equipment speed. The opening correction range is superimposed with the initial valve opening to obtain the initial target opening. If the initial target speed is greater than the preset safe speed limit, then the safe speed limit is taken as the target equipment speed; if it is not greater, then the initial target speed is taken as the target equipment speed. If the initial target opening is greater than the preset upper limit of the mechanical opening, then the upper limit of the mechanical opening is taken as the target valve opening; if it is not greater, then the initial target opening is taken as the target valve opening.

[0053] It should be noted that the compensation control command is first parsed to extract the speed correction range and opening correction range. The compensation control command is generated in the previous steps and sent to the underlying controller via the Modbus RTU protocol. The command frame is 4 bytes long, with the command header occupying 1 byte and taking the value 0xAA to identify the compensation control command type, the data variables occupying 2 bytes, and the check frame occupying 1 byte. The parsing process first verifies whether the command header is 0xAA and whether the check frame is equal to the XOR result of the command header and the data variables. After successful verification, the high 8 bits of the data variables are extracted as the speed correction range, and the low 8 bits are used as the raw encoded value of the opening correction range.

[0054] The speed correction range is stored in signed integer format, with units of r / min, and its value range covers both positive and negative directions of the compressor's adjustable speed range. The original coded value of the opening correction range is an integer from 0 to 255, which needs to be linearly mapped back to the opening percentage range from the coded range. The mapping method is to divide the coded value by 255 and then multiply it by the actual physical range of the opening correction range. The actual physical range of the opening correction range is set to a 20% distribution of the code from 0 to 255 corresponding to the full valve stroke. The unit after mapping is the opening percentage, with a positive sign indicating an increase in valve opening and a negative sign indicating a decrease in valve opening.

[0055] Next, the speed correction range is superimposed on the initial equipment speed to obtain the initial target speed. The initial equipment speed is provided by the optimization process output in step S102 or by the current speed setting value manually input from the upper-level monitoring interface, in units of r / min, retaining integer digits. The superposition operation is performed by algebraically summing the initial equipment speed value and the speed correction range, with both in the same unit of r / min. If the initial target speed is positive, it represents the normal operating range where the actual speed is greater than zero. If the superimposed value is less than or equal to zero, it is directly set to zero, the operation stops, and a speed lower limit alarm is output.

[0056] Simultaneously, the opening correction range is superimposed on the initial valve opening to obtain the initial target opening. The initial valve opening is also provided by the output of step S102 or a manually set value, in percentage of opening, rounded to one decimal place. The superposition operation involves algebraically summing the percentage value of the initial valve opening and the opening correction range, keeping the sum within the physical range of 0% to 100% without initial amplitude limiting, awaiting subsequent mechanical and safety boundary judgments.

[0057] Subsequently, a speed safety boundary check is performed. The preset safe speed limit is set based on the maximum permissible continuous operating speed specified in the equipment technical specifications provided by the compressor manufacturer. The maximum continuous speed of a gear-type centrifugal compressor is generally 105% to 110% of the rated speed, which is taken from the compressor nameplate parameters. This maximum permissible speed value is used as the safe speed limit and hard-coded into the read-only parameter storage area of ​​the underlying controller, in r / min. If the initial target speed is greater than the safe speed limit, the value of the safe speed limit is retrieved and assigned to the target equipment speed. At the same time, a speed limiting record is triggered, writing the current timestamp, the initial target speed, and the limited value into the operation log for future reference. If the initial target speed is not greater than the safe speed limit, the initial target speed is directly output as the target equipment speed.

[0058] Then, a mechanical boundary check of the valve opening is performed. The preset upper limit of the mechanical opening is set based on the mechanical stop position of the valve equipped with the electric valve positioner. The fully closed position of the valve is set to 0% opening by the limit switch contact, and the fully open position is set to 100% opening by the reverse limit switch contact. The internal firmware of the electric valve positioner automatically clamps commands exceeding the range of 0% to 100% to the boundary value. The upper limit of the mechanical opening is fixed at 100%, and the lower limit of the mechanical opening is fixed at 0%. If the initial target opening is greater than 100%, 100% is used as the target valve opening, and the opening limit information is recorded. If the initial target opening is less than 0%, 0% is used as the target valve opening. If the initial target opening falls within the range of 0% to 100%, the initial target opening is directly output as the target valve opening.

[0059] Finally, the target equipment speed and target valve opening, after being verified by both safety and mechanical boundaries, are synchronously written into the output register of the underlying controller. The speed value is written into the frequency setting register of the frequency converter, and the opening value is converted into a current signal by the 4 to 20mA analog output module and sent to the electric valve positioner, completing a complete compensation and correction closed loop. This ensures that the actual operating parameters of the compressor track the target operating conditions after efficiency optimization while meeting the equipment safety constraints.

[0060] like Figure 3 As shown in the figure, this diagram illustrates the classification of deviation exceeding the limit after feature decomposition and label generation in step S104. A comparison clearly shows that after the feature decomposition and label generation process, the exceeding-limit patterns that were originally difficult to distinguish from the continuous deviation curves (…) are now much clearer. Figure 2 All the transition points of the solid blue line in the middle are effectively identified and classified as flow overshoot indicators (red dots, concentrated in the early period when the deviation exceeds the limit in the positive direction and the response time difference is less than the preset time threshold) and response lag indicators (blue dots, distributed in the later period when the deviation exceeds the limit in the negative direction or the response time difference exceeds the threshold). At the same time, normal fluctuation points within the deviation tolerance range are filtered out and not marked. This result intuitively proves that step S104 of this scheme can effectively transform the fuzzy deviation time sequence into overshoot / lag binary indicators with clear control semantics, providing a solid classification decision basis for subsequent differential allocation of speed and valve compensation weights according to the indicator type and generation of accurate compensation control commands under energy consumption constraints.

[0061] In step S105, adjusting the compression conditions based on the target equipment speed and target valve opening to obtain the actual efficiency optimization result includes: A frequency converter drive command is generated based on the target equipment rotation speed, and a servo positioning command is generated based on the target valve opening degree. The inverter drive command and the servo positioning command are sent synchronously to the corresponding actuator to adjust the compression conditions. Continuously collect and adjust the output power sequence and the latest temperature parameters; The actual efficiency optimization result is obtained by calculating the energy consumption change rate before and after adjustment based on the output power sequence and the latest temperature parameter.

[0062] It should be noted that, firstly, the inverter drive command is generated based on the target equipment speed, and secondly, the servo positioning command is generated based on the target valve opening. The inverter drive command is generated by converting the target equipment speed value into an inverter frequency setpoint through a linear mapping. The mapping relationship is determined based on the number of pole pairs and the rated frequency of the compressor drive motor. For a motor with a rated speed of 3000 r / min corresponding to a rated frequency of 50 Hz, the frequency corresponding to each r / min of speed is approximately 0.01667 Hz (50 divided by 3000). The target speed is multiplied by this coefficient to obtain the target frequency value, in Hz, rounded to two decimal places.

[0063] The target frequency value is written into the data field of the Modbus RTU communication frame. The communication frame includes the inverter slave address (1 byte), function code 0x06 (indicating writing to a single register, 1 byte), register address (2 bytes corresponding to the internal address of the inverter frequency setting register), register value (2 bytes, calculated as the actual frequency value multiplied by 100 and rounded before transmission), and cyclic redundancy check code (2 bytes). The complete instruction frame consists of 8 bytes. (Two-pole asynchronous motor, p=1, synchronous speed n=60f / p=60×50 / 1=3000r / min).

[0064] The servo positioning command is generated by linearly mapping the target valve opening percentage value to a 4-20mA current loop output current value. The mapping relationship is 0% opening corresponds to 4mA, 100% opening corresponds to 20mA, and each 1% change in the percentage value corresponds to a 0.16mA change in current. The output current value is written to the corresponding channel register of the analog output module. The module resolution is 12 bits, with the full scale corresponding to a digital encoding value of 0 to 4095, and the 4-20mA range corresponding to a digital encoding value of 819 to 4095. After proportionally converting the target current value into a digital encoding value, it is written to the output latch via the internal bus.

[0065] Next, the inverter drive command and servo positioning command are synchronously sent to the corresponding actuator to adjust the compression condition. The synchronous transmission is achieved as follows: in the control cycle timer interrupt service routine of the underlying controller, each control cycle is fixed at 100ms. Within the same cycle, the serial transmission of the Modbus RTU communication frame and the latching of the digital encoded value of the analog output module are executed sequentially. The serial communication interface is set to RS485 half-duplex mode with a baud rate of 9600bps. After transmission, the system waits for the inverter to return a response frame to verify successful writing. If the response frame function code is abnormal or no response is received within 30ms, a communication fault retransmission is triggered, with a maximum of 3 retransmissions. The latching operation of the analog output module takes effect immediately. The output voltage-type current signal is transmitted to the analog input terminal of the electric valve positioner via a shielded twisted pair cable. The positioner's built-in PID controller drives the servo motor to rotate based on the deviation between the input current and the valve position feedback signal, moving the valve core to the target opening position. After receiving the frequency setpoint, the inverter's internal ramp function generator smoothly transitions the output frequency from the current value to the target frequency value with a ramp time of 500ms. The drive motor speed follows the output frequency according to the slip ratio. The slip ratio is approximately 1.5% to 2.5% under rated operating conditions. Finally, the steady-state deviation between the actual motor speed and the target equipment speed is within 0.5% of the rated speed.

[0066] Subsequently, the adjusted output power sequence and the latest temperature parameters are continuously collected. Output power is acquired via a power transmitter installed at the compressor drive motor inlet. The power transmitter simultaneously receives three-phase voltage and three-phase current signals, and the total active power is calculated using a three-wattmeter method. The sampling frequency is set to 10Hz, consistent with the control cycle of the underlying controller, and the output unit is kW with two decimal places. The acquired active power values ​​are stored sequentially in a circular buffer with a buffer length of 300 sampling points, covering the data of the most recent 30 seconds, forming the output power sequence. The latest temperature parameters include the compressor inlet temperature and the heat storage temperature, acquired respectively via the PT100 platinum resistance temperature sensor and the K-type thermocouple installed in step S101. Temperature sensor data is read by the data acquisition unit at a 1Hz frequency. After each reading, the latest temperature value is stored in the temperature parameter buffer, which retains the two most recently read temperature values ​​for subsequent calculations. Every 100ms control cycle, all 300 active power samples from the past 30 seconds are read from the circular buffer, and the latest temperature reading is read from the temperature parameter buffer. These are used together as the data input for this calculation.

[0067] Then, the energy consumption change rate before and after adjustment is calculated based on the output power sequence and the latest temperature parameters to obtain the actual efficiency optimization result. The baseline energy consumption data before adjustment has been collected and stored before the compensation control command is sent to the underlying controller in the previous step. The baseline energy consumption data includes the average active power calculated from the output power sequence within 30 seconds before adjustment, the latest compressor inlet temperature and heat storage temperature before adjustment, and the energy consumption value per unit flow rate before adjustment calculated and output by step S102. When calculating the energy consumption after adjustment, the arithmetic mean of 300 active power samples in the current annular buffer over the past 30 seconds is taken as the average active power after adjustment, in kW. The average active power after adjustment is divided by the actual gas mass flow rate at the current moment. The current mass flow rate is read from the instantaneous value of the latest second from the thermal mass flow meter used in step S102 to calculate the energy consumption per unit flow rate after adjustment, in kJ per kg. The energy consumption change rate is calculated by subtracting the unit flow energy consumption before adjustment from the adjusted unit flow energy consumption, dividing the difference by the unit flow energy consumption before adjustment, and multiplying by 100%. The result is a percentage value. A negative percentage value indicates a decrease in energy consumption, meaning the efficiency optimization is effective; a positive percentage value indicates an increase in energy consumption. The absolute value of the percentage value reflects the magnitude of the efficiency optimization. The adjusted average active power, adjusted unit flow energy consumption, energy consumption change rate, and the change in isentropic efficiency before and after adjustment are encapsulated into a single result structure as the actual efficiency optimization result. This structure is written to the operation log database and displayed in real-time on the monitoring interface. If the energy consumption change rate is negative and its absolute value exceeds 2% for three consecutive calculation cycles, the current target equipment speed and target valve opening are updated as new optimization baseline states in the optimization model of step S102, completing one closed-loop iteration of online optimization of the compressed air energy storage system efficiency.

[0068] In summary, this invention discloses an efficiency optimization method for compressed air energy storage systems, which can achieve dynamic efficiency optimization and precise operating condition adjustment of compressed air energy storage systems, meeting the core requirements of improving both flow response accuracy and energy conversion efficiency.

[0069] It should be noted that the efficiency optimization system for compressed air energy storage systems provided in this embodiment of the invention is used to execute all the process steps of the efficiency optimization method for compressed air energy storage systems in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0070] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. An efficiency optimization method for compressed air energy storage systems, characterized in that, include: The system acquires real-time gas storage pressure, compressor inlet temperature, thermal storage temperature, and grid dispatch commands; extracts the pressure change characteristics of the real-time gas storage pressure and performs noise reduction processing to obtain smooth state characteristics. Based on the smooth state characteristics, the compressor inlet temperature and the heat storage temperature, the isentropic efficiency and energy consumption per unit flow are calculated to obtain the real-time efficiency index. Then, optimization processing is performed based on the real-time efficiency index to obtain the initial equipment speed and the initial valve opening. The real-time flow response corresponding to the initial equipment rotation speed and initial valve opening is collected, the time change gradient is calculated to obtain the flow dynamic characteristics, and the adjustment deviation between the flow dynamic characteristics and the target flow gradient is calculated. If the adjustment deviation is greater than the deviation tolerance range, a compensation control command is calculated under the constraint that the estimated energy consumption after compensation does not exceed a preset multiple. The compensation control command is used to perform correction processing on the initial equipment speed and the initial valve opening to obtain the target equipment speed and the target valve opening. If not greater than, then the initial equipment speed and the initial valve opening are taken as the target equipment speed and the target valve opening; The compression conditions are adjusted based on the target equipment speed and target valve opening to obtain the actual efficiency optimization results.

2. The efficiency optimization method for compressed air energy storage systems according to claim 1, characterized in that, The process of acquiring real-time gas storage pressure, compressor inlet temperature, thermal storage temperature, and grid dispatch commands, extracting the pressure change characteristics of the real-time gas storage pressure, and performing noise reduction processing to obtain smooth state characteristics includes: The system acquires real-time gas storage pressure, compressor inlet temperature, thermal storage temperature, and grid dispatch commands, and performs time-series matching processing to obtain a synchronized time series. The real-time gas storage pressure in the synchronous time series is calculated using first-order difference to obtain the pressure change sequence, and pressure fluctuation characteristics are extracted based on the pressure change sequence. The initial operating data is obtained by concatenating the pressure fluctuation characteristics, the power grid dispatch instructions in the synchronous time series, the compressor inlet temperature and the thermal storage temperature execution characteristics. A Kalman filter model is constructed based on the initial running data, and the noise covariance matrix of the Kalman filter model is extracted. The initial running data is time-updated using the noise covariance matrix to obtain a priori state estimation sequence; a gain adjustment matrix is ​​calculated based on the priori state estimation sequence; and a measurement update process is performed on the priori state estimation sequence using the gain adjustment matrix to obtain a posterior state estimation sequence. High-frequency interference filtering is performed on the posterior state estimation sequence to obtain smooth state features.

3. The efficiency optimization method for compressed air energy storage systems according to claim 1, characterized in that, The process of calculating isentropic efficiency and energy consumption per unit flow rate based on the smooth state characteristics, the compressor inlet temperature, and the heat storage temperature to obtain real-time efficiency indicators includes: The current actual compression ratio is analyzed based on the smooth state characteristics; The ideal isentropic compression work is calculated by combining the compressor inlet temperature, the heat storage temperature, and the actual compression ratio. The actual shaft work consumed by the compressor is obtained, and the ratio of the ideal isentropic compression work to the actual shaft work consumed is calculated to obtain the isentropic efficiency. The actual gas mass flow rate is obtained, and the ratio of the actual shaft work consumed to the actual gas mass flow rate is calculated to obtain the energy consumption per unit flow rate. The isentropic efficiency and the energy consumption per unit flow are used as real-time efficiency indicators.

4. The efficiency optimization method for compressed air energy storage systems according to claim 1, characterized in that, The step of performing optimization processing based on the real-time efficiency index to obtain the initial equipment speed and initial valve opening includes: Extract the steady-state and transient components from the smooth state features; The steady-state component, the transient component, and the real-time efficiency index are input into a preset polynomial regression fitting model to perform fitting calculations and establish a multidimensional nonlinear mapping equation between efficiency and flow output. Based on the multidimensional nonlinear mapping equation, under the boundary condition of ensuring that the flow output matches the power grid dispatch command, analytical optimization is performed with the goal of minimizing the energy consumption per unit flow of the system to obtain the initial equipment speed and the initial valve opening.

5. The efficiency optimization method for a compressed air energy storage system according to claim 1, characterized in that, The process involves acquiring the real-time flow response corresponding to the initial equipment rotation speed and initial valve opening, calculating the time-varying gradient to obtain the flow dynamic characteristics, and calculating the adjustment deviation between the flow dynamic characteristics and the target flow gradient, including: Based on the initial equipment rotation speed and the initial valve opening, a drive command is generated and sent to the underlying controller, and the real-time flow response of the compressor exhaust port is collected; The real-time traffic response is subjected to data cleaning processing to obtain an effective traffic sequence; Extract adjacent time nodes from the effective traffic sequence; Calculate the ratio of the flow difference between adjacent time nodes to the time interval to obtain the time change gradient; Perform feature aggregation processing on the time-varying gradient to obtain dynamic traffic features; The difference between the dynamic characteristics of the flow and the target flow change gradient corresponding to the power grid dispatch command is calculated to obtain the adjustment deviation.

6. The efficiency optimization method for a compressed air energy storage system according to claim 1, characterized in that, If the adjustment deviation is greater than the deviation tolerance range, a compensation control command is calculated under the constraint that the estimated energy consumption after compensation does not exceed a preset multiple, including: The adjustment deviation is subjected to feature decomposition processing to obtain the over-limit direction feature and response time difference; If the over-limit direction feature is positive and the response time difference is less than a preset time threshold, a flow overshoot flag is generated; if the over-limit direction feature is negative or the response time difference is not less than the preset time threshold, a response lag flag is generated. Based on the flow overshoot flag or the response lag flag, the adjustment deviation is multiplied by the current pressure to obtain the basic compensation base. Based on the flow overshoot flag or the response hysteresis flag, the corresponding weights are assigned to the basic compensation base to obtain the primary speed compensation amount and the primary valve compensation amount. Energy consumption simulation calculations are performed using the primary speed compensation amount and the primary valve compensation amount to obtain the estimated energy consumption per unit flow rate. If the estimated unit flow energy consumption does not exceed the preset multiple before adjustment, then the primary speed compensation amount and the primary valve compensation amount are used as the speed compensation amount and valve compensation amount; if they exceed, then the primary speed compensation amount and the primary valve compensation amount are subjected to proportional attenuation processing to obtain the speed compensation amount and valve compensation amount that meet the constraint conditions. The speed compensation amount and the valve compensation amount are spliced ​​together to obtain a compensation control command.

7. The efficiency optimization method for a compressed air energy storage system according to claim 1, characterized in that, The step of using the compensation control command to correct the initial equipment speed and initial valve opening to obtain the target equipment speed and target valve opening includes: The compensation control command is parsed to extract the speed correction range and opening correction range; The initial target speed is obtained by superimposing the speed correction range with the initial equipment speed. The opening correction range is superimposed with the initial valve opening to obtain the initial target opening. If the initial target speed is greater than the preset safe speed limit, then the safe speed limit is taken as the target equipment speed; if it is not greater, then the initial target speed is taken as the target equipment speed. If the initial target opening is greater than the preset upper limit of the mechanical opening, then the upper limit of the mechanical opening is taken as the target valve opening; if it is not greater, then the initial target opening is taken as the target valve opening.

8. The efficiency optimization method for a compressed air energy storage system according to claim 1, characterized in that, The step of adjusting the compression conditions based on the target equipment speed and target valve opening to obtain the actual efficiency optimization result includes: A frequency converter drive command is generated based on the target equipment rotation speed, and a servo positioning command is generated based on the target valve opening degree. The inverter drive command and the servo positioning command are synchronously sent to the corresponding actuator to adjust the compression conditions. Continuously collect and adjust the output power sequence and the latest temperature parameters; The actual efficiency optimization result is obtained by calculating the energy consumption change rate before and after adjustment based on the output power sequence and the latest temperature parameter.

9. An efficiency optimization system for a compressed air energy storage system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the efficiency optimization method for a compressed air energy storage system as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the efficiency optimization method for a compressed air energy storage system as described in any one of claims 1 to 8.