Electric heating gas energy storage synergetic charge energy state disturbance suppression optimization method

Abstract

The invention provides an electric heating gas energy storage collaborative charge state disturbance suppression optimization method, and relates to the technical field of information, and the method comprises the steps: carrying out the filtering processing of a second-level disturbance signal in the real-time operation data of a power grid through an adaptive Kalman filtering algorithm, so as to improve the signal-to-noise ratio; performing time-frequency domain feature extraction on the filtered disturbance signal by using a wavelet analysis method, and judging the type and severity of the disturbance signal in combination with a preset disturbance mode library; and according to the type and characteristics of the disturbance signal. According to the method, an electric energy storage collaborative optimization control model is constructed by adopting a robust optimization theory, the uncertainty of disturbance signals and electric energy storage characteristics is considered, an optimal power instruction is solved, in the control process, the state of an electric energy storage unit is dynamically evaluated, a strategy is adjusted, online optimization is carried out according to power grid index changes, the control effect is continuously improved, and the control efficiency is improved. Through cooperative optimization control of electric energy storage, second-level disturbance of the power grid is effectively suppressed, and safe and stable operation of the power grid is guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for optimizing charge state disturbance suppression in electric, thermal and gas energy storage synergy. Background Art

[0002] There are various types of second-level disturbances in the power grid, such as sudden load changes and fluctuations in the output of new energy sources. These disturbances will cause rapid changes in key indicators such as grid frequency and voltage, affecting the safe and stable operation of the grid. Traditional grid regulation methods such as generator frequency regulation and reactive power compensation have problems such as slow response speed and insufficient regulation capacity, making it difficult to effectively suppress the impact of second-level disturbances. Therefore, it is urgent to study fast-response collaborative control strategies to give full play to the advantages of flexible and controllable electric energy storage and achieve rapid suppression of second-level disturbances.

[0003] However, the participation of electric energy storage in the optimization control of second-level disturbance suppression faces a series of technical challenges. First, second-level disturbances are highly random and short-lasting, making it difficult to accurately capture and characterize disturbance signals. Secondly, electric energy storage itself has power and capacity constraints, and the dynamic characteristics of different types of electric energy storage vary significantly, making it difficult to achieve collaborative optimization control of multiple types of electric energy storage. In addition, in the actual power grid operation environment, the disturbance signal is easily contaminated by measurement noise, the dynamic response characteristics of electric energy storage may also change, and the robustness of the control strategy needs to be improved. In summary, the study of adaptive and robust collaborative optimization control strategies for electric energy storage is of great significance for improving the power grid's ability to resist second-level disturbances and ensuring the safe and stable operation of the power grid. Summary of the invention

[0004] The present invention provides an optimization of charge state disturbance suppression for coordinated electric, thermal and gas energy storage, which mainly includes:

[0005] An adaptive Kalman filter algorithm is used to filter the second-level disturbance signals in the real-time operation data of the power grid to improve the signal-to-noise ratio; the wavelet analysis method is used to extract the time-frequency domain features of the filtered disturbance signal, and the type and severity of the disturbance signal are judged in combination with the preset disturbance pattern library; according to the type and characteristics of the disturbance signal, energy storage units with complementary dynamic characteristics are selected from the energy storage group, and an online identification model of the dynamic characteristics of energy storage is established; the robust optimization theory is used to construct an energy storage collaborative optimization control model, and the control strategy of the energy storage group is adjusted in real time to ensure that the charging and discharging behavior of the energy storage group will not cause secondary disturbances to the power grid; the execution effect of the energy storage collaborative optimization control strategy is evaluated in real time, and the control strategy is optimized online according to the requirements of safe and stable operation of the power grid.

[0006] Furthermore, the adaptive Kalman filtering algorithm is used to filter the second-level disturbance signals in the real-time operation data of the power grid, including: obtaining the real-time operation data of the power grid, estimating the initial measurement noise covariance matrix and the process noise covariance matrix by the least squares method, calculating the difference between the estimated value and the measured value of the disturbance signal in real time to obtain the residual sequence, and adjusting the values ​​of the measurement noise covariance matrix and the process noise covariance matrix in real time according to the mean and variance of the residual sequence.

[0007] Furthermore, the use of wavelet analysis method to extract time-frequency domain features of the filtered disturbance signal includes: performing wavelet transform on the filtered disturbance signal to obtain an energy distribution matrix of the signal in the time-frequency domain, extracting a time-frequency region with relatively concentrated energy, determining the main frequency components in the region, and constructing a multidimensional feature vector.

[0008] Furthermore, selecting electric energy storage units with complementary dynamic characteristics from the electric energy storage group according to the type and characteristics of the disturbance signal includes: selecting the optimal combination from the electric energy storage group using a multi-objective optimization algorithm such as a particle swarm optimization or a genetic algorithm according to dynamic characteristic parameters such as response speed, duration and power capacity of each unit in the electric energy storage group to achieve complementary dynamic characteristics.

[0009] Furthermore, the use of robust optimization theory to construct an electric energy storage collaborative optimization control model includes: establishing a mathematical description that takes uncertainty into account by analyzing historical operating data of the electric energy storage system, using the established electric energy storage dynamic model containing uncertainty as a constraint condition, and using multi-objective optimization algorithms such as NSGA-II to solve multi-objective optimization problems that take uncertainty into account, thereby obtaining a Pareto optimal solution set for collaborative optimization control of the electric energy storage system.

[0010] Furthermore, the real-time evaluation of the execution effect of the electric energy storage collaborative optimization control strategy includes: obtaining real-time operation data of the power grid, including key indicators such as power grid frequency and voltage, judging whether the power grid is in a safe and stable state according to preset thresholds, and using a fuzzy comprehensive evaluation method to calculate the control effect score. If the control effect score is lower than the preset threshold, the adaptive weight adjustment mechanism is triggered.

[0011] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0012] The present invention discloses a method for optimizing the suppression of disturbances in the energy state of electric, thermal and gas energy storage synergistically. The method first uses adaptive Kalman filtering and wavelet analysis to process and extract features of second-level disturbance signals in the real-time operation data of the power grid to determine the type and severity of the disturbance. Then, according to the disturbance characteristics, an electric energy storage unit with complementary dynamic characteristics is selected from the electric energy storage group, and an online identification model of its dynamic characteristics is established. The robust optimization theory is used to construct an electric energy storage collaborative optimization control model, and the uncertainty of the disturbance signal and the electric energy storage characteristics is considered to solve the optimal power instruction. During the control process, the state of the electric energy storage unit is dynamically evaluated and the strategy is adjusted. At the same time, online optimization is performed according to the changes in the power grid indicators to continuously improve the control effect. The present invention effectively suppresses second-level disturbances in the power grid and ensures the safe and stable operation of the power grid through the collaborative optimization control of electric energy storage. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a flow chart of the method for optimizing the suppression of charge state disturbances of the electric, thermal and gas energy storage collaboration of the present invention. DETAILED DESCRIPTION

[0014] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0015] like Figure 1 In this embodiment, the optimization of charge state disturbance suppression of electric, thermal and gas energy storage synergy can specifically include:

[0016] Step S101, based on the real-time operation data of the power grid, an adaptive Kalman filter algorithm is used to filter the second-level disturbance signal to remove the influence of measurement noise and improve the signal-to-noise ratio of the disturbance signal. The algorithm achieves adaptation to different noise environments by adjusting the measurement noise covariance matrix and the process noise covariance matrix in real time.

[0017] The real-time operation data of the power grid is obtained, and the adaptive Kalman filter algorithm is used to filter the second-level disturbance signal. According to the power grid operation data, the values ​​of the initial measurement noise covariance matrix R and the process noise covariance matrix Q are estimated using the least squares method as the input parameters of the Kalman filter algorithm. During the filtering process, the difference between the estimated value and the measured value of the disturbance signal is calculated in real time to obtain the residual sequence. The mean and variance of the residual sequence are calculated using a sliding window. If the residual mean exceeds the preset threshold or the residual variance exceeds the preset threshold, it is judged that the noise environment has changed. When the noise environment changes, the maximum likelihood estimation method is used to re-estimate the values ​​of the measurement noise covariance matrix R and the process noise covariance matrix Q based on the residual mean and variance. The updated R and Q are used as new input parameters to update the state equation and observation equation of the Kalman filter algorithm. Repeat the process of residual calculation, noise environment judgment and parameter adaptive adjustment until the residual mean and variance in N consecutive windows meet the preset convergence conditions, that is, the residual mean is less than ε1 and the residual variance is less than ε2. The filtering process is considered to converge, and the estimated value of the disturbance signal after removing the influence of measurement noise is obtained. Calculate the signal-to-noise ratio of the disturbance signal before and after filtering, and evaluate the effect of the adaptive Kalman filter algorithm on improving the signal-to-noise ratio. If the improvement of the signal-to-noise ratio does not meet expectations, adjust the ε1 and ε2 values ​​in the convergence conditions and the size of the sliding window according to the improvement of the signal-to-noise ratio, and further optimize the parameter settings of the filtering algorithm. After the above adaptively optimized Kalman filter processing, the measurement noise in the power grid disturbance signal can be effectively removed, the signal-to-noise ratio of the disturbance signal can be improved, and high-quality data support can be provided for subsequent power grid disturbance analysis and fault diagnosis.

[0018] Exemplarily, the real-time operation data of the power grid contains various signals, such as voltage, current, frequency, etc. These signals are often mixed with various noises, among which the second-level disturbance signal is the key to the stability analysis of the power grid. In order to extract clean disturbance signals, filtering processing is required. The adaptive Kalman filter algorithm is an effective filtering method, which can automatically adjust the filtering parameters according to the changes in the noise environment to obtain better filtering effects. The core of the Kalman filter algorithm is the state equation and the observation equation. The state equation describes the change law of the disturbance signal over time, and the observation equation describes the relationship between the measured value and the disturbance signal. Both equations contain noise terms, which are represented by the process noise covariance matrix Q and the measurement noise covariance matrix R respectively. The initial values ​​of these two matrices can be estimated by the least squares method. For example, the power grid operation data for a period of time can be collected, the change curve of the disturbance signal can be fitted by the least squares method, and the variance of the fitting residual can be calculated to estimate the initial values ​​of Q and R. Assuming that the measured voltage disturbance signal sequence is 01, 02, 03, 02, 01, and a straight line y=02 is obtained by least squares fitting, then the residual sequence is 01, 0, 01, 0, -01, and the initial value of R can be set to the variance of the residual sequence. The initial value of Q can be estimated according to the severity of the change of the disturbance signal. The more severe the change, the larger the Q value. The Kalman filter algorithm predicts the next value of the disturbance signal according to the state equation, and fuses the predicted value with the actual measured value according to the observation equation to obtain the final estimated value. In this process, the residual sequence, that is, the difference between the estimated value and the measured value, is the key indicator for judging whether the noise environment has changed. Assuming that in a sliding window, the calculated residual sequence is 01, 02, 03, 04, 05, if the preset residual mean threshold is 02 and the residual variance threshold is 0005, then the residual mean of the current window is 03 and the residual variance is 0002. Since the residual mean exceeds the threshold, it is judged that the noise environment has changed. When the noise environment changes, the values ​​of Q and R need to be re-estimated. The maximum likelihood estimation method is a commonly used parameter estimation method. It estimates Q and R by finding the parameter value that maximizes the probability of the observed data. For example, if the variance of the residual sequence increases, it means that the measurement noise increases and the R value should increase accordingly. Assuming that the re-estimated R value is 001 and the Q value is 1, the updated R and Q are substituted into the Kalman filter algorithm to update the state equation and observation equation. Repeat the process of residual calculation, noise environment judgment and parameter adaptive adjustment until the filtering process converges. The convergence condition can be set to the residual mean in N consecutive windows is less than ε1 and the residual variance is less than ε2. For example, N=10, ε1=01, ε2=0001 can be set. This means that the residual mean in 10 consecutive windows is less than 01 and the residual variance is less than 0001, then the filtering process is considered to converge. The signal-to-noise ratio of the disturbance signal before and after filtering is an important indicator for evaluating the filtering effect.The improvement of the signal-to-noise ratio indicates that the filtering algorithm has effectively removed the noise. Assuming that the signal-to-noise ratio of the disturbance signal before filtering is 10dB, and the signal-to-noise ratio after filtering is increased to 20dB, it means that the filtering algorithm has effectively improved the signal-to-noise ratio. If the improvement of the signal-to-noise ratio does not meet expectations, it is necessary to adjust the ε1 and ε2 values ​​in the convergence conditions and the size of the sliding window. For example, the values ​​of ε1 and ε2 can be reduced, or the size of the sliding window can be increased to improve the filtering accuracy. Through the adaptive Kalman filtering algorithm, the measurement noise in the power grid disturbance signal can be effectively removed, the signal-to-noise ratio of the disturbance signal can be improved, and high-quality data support can be provided for subsequent power grid disturbance analysis and fault diagnosis. For example, the disturbance signal after filtering can be used to analyze the stability of the power grid, or to identify the type and location of the power grid fault.

[0019] Step S102, the filtered disturbance signal is subjected to feature extraction using a wavelet analysis method to obtain the time-frequency domain features of the second-level disturbance signal. Combined with the preset disturbance pattern library, the type and severity of the disturbance signal are determined. The time-frequency domain features include energy distribution, frequency component changes, and duration.

[0020] Obtain the disturbance signal after filtering, perform wavelet transform on the signal, and obtain the energy distribution matrix of the signal in the time-frequency domain. Wavelet transform can effectively extract the time-frequency characteristics of the signal, providing a basis for subsequent feature analysis. According to the energy distribution matrix, extract the time-frequency area with relatively concentrated energy, determine the main frequency components in the area, and obtain the curve of the frequency component changing with time. By analyzing the morphological characteristics of the frequency component change curve, including the rising rate, the falling rate, the duration, etc., a multidimensional feature vector is constructed. The K-means clustering algorithm is used to classify the feature vectors, and the Euclidean distance is calculated with the feature vectors of each mode in the preset disturbance mode library according to the coordinates of the center point of each category to obtain the similarity. If the Euclidean distance between the feature vector of a certain category and a certain mode in the library is less than the preset threshold, it is judged that the current disturbance signal belongs to the type corresponding to the mode. According to the parameters such as the energy size of the disturbance signal and the amplitude of the frequency component, combined with the energy and amplitude thresholds obtained based on historical data statistics, the severity of the disturbance is evaluated to obtain qualitative severity indicators, such as "mild", "moderate", "severe", etc. The type and severity index of the disturbance signal are output for subsequent control decision and fault diagnosis modules. The control decision module adjusts the control strategy according to the type and severity of the disturbance to reduce the impact of the disturbance on the system. The fault diagnosis module determines the possible type of fault based on the disturbance information and prompts targeted inspection and maintenance. By real-time monitoring and responding to disturbances, the stability and reliability of the system can be improved.

[0021] Exemplarily, after the disturbance signal passes through the adaptive Kalman filter, a relatively pure disturbance signal can be obtained. In order to further analyze the characteristics of the disturbance signal, it is necessary to perform time-frequency analysis on it. Wavelet transform is an effective time-frequency analysis method, which can decompose the signal into a superposition of different frequency components and display the energy distribution of each frequency component in different time periods. Assume that the filtered voltage disturbance signal is a decaying oscillation waveform. After performing wavelet transform on it, an energy distribution matrix can be obtained. The rows of the matrix represent time, and the columns represent frequency. The elements in the matrix represent the energy of the signal at this time point and the frequency component. For example, the upper left corner element of the matrix represents the energy of the signal on the low-frequency component at the initial moment, and the lower right corner element represents the energy of the signal on the high-frequency component at the end moment. By observing the energy distribution matrix, it can be found that the energy is mainly concentrated in a specific time-frequency area. For example, the energy may be concentrated in the low-frequency area at the initial moment, then gradually move to the high-frequency area, and finally attenuate and disappear. This shows that the frequency component of the disturbance signal has changed over time. According to the energy distribution matrix, the time-frequency area with relatively concentrated energy can be extracted, and the main frequency components in the area can be determined. For example, an energy threshold can be set to extract the time-frequency points with energy greater than the threshold, and then the frequency values ​​corresponding to these time-frequency points are counted to obtain a curve of the frequency component changing over time. Assume that the frequency component change curve shows a trend of first rising rapidly and then falling slowly. Feature parameters such as the rising rate, the falling rate, and the duration can be extracted to construct a multidimensional feature vector. For example, the rising rate can be set as the first element of the feature vector, the falling rate can be set as the second element, and the duration can be set as the third element. In order to classify the disturbance signal, the K-means clustering algorithm can be used to perform cluster analysis on the feature vector. The K-means algorithm divides the feature vector into K categories, each with a center point. Assume that the preset disturbance pattern library contains three types of disturbances: A, B, and C. Each type corresponds to a feature vector. For example, the feature vector of type A disturbance is (1, 5, 2), the feature vector of type B disturbance is (5, 1, 1), and the feature vector of type C disturbance is (2, 1, 5). The similarity can be obtained by calculating the Euclidean distance between the center point coordinates of each category and the feature vectors of each mode in the disturbance pattern library. If the Euclidean distance between the feature vector of a category and a mode in the library is less than the preset threshold, the current disturbance signal is judged to belong to the type corresponding to the mode. For example, assuming that the center point coordinates of a category are (9, 6, 8), and the Euclidean distance with the class A disturbance is 2, which is less than the preset threshold 3, then the disturbance signal is judged to belong to the class A disturbance. In addition to the type of disturbance, the severity of the disturbance also needs to be evaluated. The severity of the disturbance can be evaluated based on parameters such as the energy size of the disturbance signal and the amplitude of the frequency component, combined with the energy and amplitude thresholds obtained based on historical data statistics.For example, if the energy of the disturbance signal exceeds the preset energy threshold, and the amplitude of the frequency component exceeds the preset amplitude threshold, the disturbance is judged to be a "severe" disturbance. Otherwise, if only one of the thresholds is exceeded, it is judged to be a "moderate" disturbance. If neither threshold is exceeded, it is judged to be a "mild" disturbance. Finally, the type and severity index of the disturbance signal are output for use by subsequent control decision and fault diagnosis modules. For example, if a Class A "severe" disturbance is detected, the control decision module may take emergency control measures to reduce the impact of the disturbance on the system. The fault diagnosis module may prompt for inspection and maintenance related to Class A faults. This can effectively improve the stability and reliability of the system.

[0022] Step S103, for different types of second-level disturbances determined, select energy storage units with complementary dynamic characteristics from the energy storage group. Complementary dynamic characteristics refer to the complementarity of different energy storage units in response speed, duration and power capacity. An online identification model of the dynamic characteristics of energy storage is established, and the model is dynamically corrected through real-time operation data to obtain an accurate description of the dynamic characteristics of energy storage.

[0023] Obtain the second-level disturbance information in the power grid, and determine the power support and response time requirements that the energy storage unit needs to provide according to the type and characteristics of the disturbance. According to the dynamic characteristic parameters of each unit in the energy storage group, including response speed, duration and power capacity, combined with the power support and response time requirements determined in the first step, establish the dynamic characteristic model of the energy storage group. Collect the operating data of the energy storage unit in real time, including power output, energy state, etc., as the input of the online identification model. Through online identification algorithms such as recursive least squares or Kalman filtering, the dynamic characteristic model of the energy storage is corrected in real time to obtain an accurate description of the dynamic characteristics. According to the corrected dynamic characteristic model of the energy storage, evaluate the response capability and complementarity of each energy storage unit to different types of second-level disturbances. Use multi-objective optimization algorithms such as particle swarm optimization or genetic algorithm, comprehensively consider factors such as response speed, duration and power capacity, select the optimal combination from the energy storage group, and achieve complementary dynamic characteristics. The optimized combination of energy storage units is sent to the energy storage group control system. Based on the real-time second-level disturbance information obtained in real time, the energy storage units are dispatched in real time to provide the required power support and maintain the stable operation of the power grid.

[0024] For example, grid disturbance is one of the challenges faced by the stable operation of the power system. Second-level disturbances, such as voltage fluctuations and frequency fluctuations, pose a threat to the stability and reliability of the power grid. The electric energy storage system, due to its fast response and flexible power regulation capabilities, has become an effective means to deal with second-level disturbances. How to coordinate the electric energy storage unit so that its dynamic characteristics complement the disturbance characteristics and effectively suppress the disturbance is a hot topic in current research. First, it is necessary to obtain the second-level disturbance information in the power grid. This can be achieved by deploying various sensors in the power grid, such as voltage transformers, current transformers, etc., to monitor the operating status of the power grid in real time. Assume that the voltage drops by 1pu at a certain moment and lasts for about 5 seconds. According to the type and characteristics of the disturbance, determine the power support and response time requirements provided by the electric energy storage unit. The voltage drop indicates that the electric energy storage unit needs to inject active power to support the voltage. According to the voltage drop and duration, it is preliminarily estimated that 1MW of power support is required and the response time is required to be within 1 second. Then, it is necessary to establish a dynamic characteristic model of the electric energy storage group. This requires consideration of the dynamic characteristic parameters of each unit in the energy storage group, including response speed, duration, and power capacity. Assume that the energy storage group consists of multiple units of different types, such as lithium-ion batteries, supercapacitors, etc. The response speed of lithium-ion batteries is slow, about 2 seconds, the continuous discharge time is long, up to several hours, and the power capacity is large, up to hundreds of MW. The response speed of supercapacitors is very fast, about 0.1 seconds, the continuous discharge time is short, about several minutes, and the power capacity is small, about tens of MW. Combined with the power support and response time requirements determined in the first step, a dynamic characteristic model of the energy storage group is established. This model can describe the characteristics of the output power of different types of energy storage units changing over time. In order to ensure the accuracy of the model, the model needs to be corrected in real time. The operating data of the energy storage unit, including power output, energy state, etc., is collected in real time as the input of the online identification model. Assume that the real-time power output data of a lithium-ion battery unit is 10MW and the energy state is 80%. Through online identification algorithms such as recursive least squares or Kalman filtering, the dynamic characteristic model of the energy storage is corrected in real time to obtain an accurate description of the dynamic characteristics. For example, the response speed and power capacity parameters of lithium-ion battery cells can be corrected according to real-time data. The corrected dynamic characteristic model of electric energy storage can be used to evaluate the response capability and complementarity of each electric energy storage unit to different types of second-level disturbances. For example, for the voltage sag disturbance mentioned above, the response capability of lithium-ion batteries and supercapacitors can be evaluated. Supercapacitors can provide power support quickly due to their fast response speed, but the duration is short. Lithium-ion batteries respond slowly, but can provide power support for a longer period of time. The complementarity between the two can ensure that sufficient power support is provided quickly at the beginning of the disturbance and maintain stable power output during the duration of the disturbance. In order to find the optimal combination of electric energy storage units, a multi-objective optimization algorithm is required.By using multi-objective optimization algorithms such as particle swarm optimization or genetic algorithm, factors such as response speed, duration and power capacity are comprehensively considered to select the optimal combination from the energy storage group to achieve complementary dynamic characteristics. For example, response speed, duration and power capacity can be used as optimization targets to search for the optimal energy storage unit combination that can meet the three targets at the same time. Assume that the optimization result is: the optimal combination of 2 supercapacitors and 1 lithium-ion battery can provide 1MW of power support, with a duration of 1 second and a response time of 0.5 seconds. Finally, the optimized energy storage unit combination scheme is sent to the energy storage group control system, and the energy storage unit is dispatched in real time according to the real-time second-level disturbance information obtained, to provide the required power support and maintain the stable operation of the power grid. For example, when a voltage sag disturbance is detected, the control system will immediately dispatch 2 supercapacitors and 1 lithium-ion battery, and quickly inject active power according to the preset control strategy, so as to control the voltage drop within the allowable range and ensure the stable operation of the power grid.

[0025] Step S104, using robust optimization theory to construct an energy storage collaborative optimization control model, introducing an uncertain set to describe the randomness of the disturbance signal and the uncertainty of the dynamic characteristics of the energy storage. The uncertain set is in the form of an ellipsoid set or a polyhedron set. The dynamic characteristics model of the energy storage obtained in the third step is used as a constraint condition, and the optimal power instruction of each energy storage unit is obtained by solving a multi-objective optimization problem considering uncertainty.

[0026] The real-time operation data of the electric energy storage system is obtained, including the state parameters of the energy storage unit and the grid disturbance signal, and the dynamic characteristic model of the electric energy storage system is constructed based on the obtained data. By analyzing the historical operation data of the electric energy storage system, the uncertainty distribution law of the disturbance signal and the dynamic characteristics of the electric energy storage is summarized, and a mathematical description considering uncertainty is established in the form of an ellipsoid set or a polyhedron set, and it is combined with the dynamic characteristic model of the electric energy storage to obtain a dynamic model of the electric energy storage system containing uncertainty. The established dynamic model of the electric energy storage containing uncertainty is used as a constraint condition, and the objective function of the multi-objective optimization problem is constructed in combination with the operation objectives of the electric energy storage system, such as smoothing power fluctuations and reducing operating costs. Multi-objective optimization algorithms such as NSGA-II are used to solve the multi-objective optimization problem considering uncertainty, and the Pareto optimal solution set of the collaborative optimization control of the electric energy storage system is obtained. According to the grid dispatching requirements and the operation status of the electric energy storage system, the optimal power instruction is selected from the Pareto optimal solution set, and the optimal power instruction is sent to the control system of each electric energy storage unit through the interface of the electric energy storage management system. The control system of each energy storage unit adjusts the operating state of the energy storage unit according to the received power command to realize dynamic optimization control of the energy storage system. The operating state of the energy storage system and the disturbance signal of the power grid are continuously monitored. When the operating state or disturbance signal is monitored to have a significant change, the update of the dynamic characteristic model and uncertainty description of the energy storage is triggered. The model is adaptively adjusted online through incremental learning, and the multi-objective optimization solution is re-performed to generate the updated optimal control command, so as to realize the adaptive update of the coordinated optimization control of the energy storage.

[0027] For example, in the power system, the dynamic characteristic model of the electric energy storage system is one of the key technologies. For example, when the frequency of the power grid fluctuates, the electric energy storage system needs to respond quickly to help stabilize the frequency of the power grid by adjusting the output power. First, the operation data of the electric energy storage system and the disturbance signal of the power grid are collected through real-time monitoring equipment such as voltage transformers and current transformers. These data include parameters such as the charging and discharging state, voltage, and current of the electric energy storage unit. Based on these real-time data, a dynamic characteristic model of the electric energy storage system can be constructed. This model can describe the behavior of the electric energy storage unit under different operating conditions, such as charging and discharging rates, response time, etc. For example, if the charging rate of a lithium-ion battery cell is 1C, it means that it can be fully charged in one hour; while its discharge rate may be 2C, that is, it can be fully discharged in half an hour. In order to deal with uncertainties in actual operation, such as performance fluctuations of the electric energy storage unit or sudden changes in the load of the power grid, ellipsoid sets or polyhedron sets can be used to describe these uncertainties. This mathematical description can help understand the performance of the electric energy storage system in the worst case and how to design control strategies to deal with these extreme situations. For example, if it is expected that the grid load may increase by 10% in a short period of time, an ellipsoid set can be designed to include all situations that may cause the output power demand of the electric energy storage system to increase by 10%. Applying this dynamic model containing uncertainty to the operating objectives of the electric energy storage system, such as smoothing power fluctuations or reducing operating costs, a multi-objective optimization problem can be constructed. The objective function of this problem may include minimizing operating costs and maximizing grid stability. Using multi-objective optimization algorithms such as NSGA-II, this problem can be solved to obtain a series of Pareto optimal solutions that represent the best trade-offs between different objectives. For example, a possible optimal solution is to increase the output power of the electric energy storage system when the grid is highly loaded to reduce the burden on the fuel power plant and thus reduce operating costs. Another solution may be to reduce the output power when the grid is under low load to extend the service life of the electric energy storage unit. According to the real-time demand of the grid and the operating status of the electric energy storage system, the power instruction that best suits the current situation can be selected from the Pareto optimal solution set. These instructions are sent to the control system of each electric energy storage unit through the electric energy storage management system interface. Each control system adjusts the operating status of the electric energy storage unit according to the received instructions to achieve dynamic optimization control of the electric energy storage system. It is necessary to continuously monitor the operating status of the energy storage system and the disturbance signals of the power grid. Once significant changes are detected, such as the performance degradation of the energy storage unit or the increase in the grid load, it is necessary to trigger the update of the dynamic characteristic model of the energy storage and the uncertainty description. The model is adaptively adjusted online through incremental learning, and the multi-objective optimization solution is re-performed to generate updated optimal control instructions. This adaptive update mechanism ensures that the energy storage system can continue to operate in the optimal state and respond to the real-time needs of the power grid.

[0028] Step S105: During the coordinated optimization control of the electric energy storage, the remaining capacity and power regulation capability of each electric energy storage unit are dynamically evaluated. Combined with the safety and stability constraints of the power grid, the control strategy of the electric energy storage group is adjusted in real time to ensure that the charging and discharging behavior of the electric energy storage group will not cause secondary disturbances to the power grid.

[0029] Obtain the real-time operating status data of each energy storage unit, including the current remaining capacity and available power regulation capability. According to the grid security and stability constraints, determine the power range and charging and discharging strategy boundaries that the energy storage group can participate in regulation. Use the particle swarm optimization algorithm to comprehensively consider factors such as the characteristics of the energy storage unit and the grid dispatching requirements to solve the optimal control strategy of the energy storage group. Monitor the operating status of the energy storage group and the changes in the grid flow in real time, and send the optimized control strategy to each energy storage unit to guide it to perform power regulation and charging and discharging operations. If it is detected that the control strategy may cause secondary disturbances in the grid, the strategy adjustment mechanism is triggered. Through rolling optimization, the control strategy of the energy storage group is dynamically updated, and a new optimal control strategy is generated and issued for execution to ensure that it is compatible with the grid security and stability constraints. Continuously track the health status and life changes of the energy storage unit. When the health index of the energy storage unit is lower than the preset threshold, adjust its capacity and power limit for participating in grid regulation to avoid excessive use and accelerated aging of the energy storage equipment.

[0030] Exemplarily, in the power system, the real-time operating status data of the electric energy storage unit, such as the current remaining capacity and the available power regulation capability, is the basis of electric energy storage management. For example, a specific electric energy storage unit may show that its remaining capacity is 80% and the current adjustable power is 500 kilowatts. These data are essential for formulating the charging and discharging strategy of the electric energy storage group. According to the safety and stability constraints of the power grid, such as frequency and voltage limits, the power range in which the electric energy storage group can participate in the regulation can be determined. For example, if the grid frequency is low, the electric energy storage group may need to increase the discharge power to provide additional power support. It is an effective method to use the particle swarm optimization algorithm to solve the optimal control strategy of the electric energy storage group. The particle swarm optimization algorithm finds the optimal solution by simulating the social behavior of bird flocks. In electric energy storage management, each "particle" represents a possible electric energy storage control strategy, including the charging and discharging power of each unit. The algorithm continuously adjusts the position of the particles (i.e., the control strategy) through an iterative process to find a strategy that minimizes the cost and meets the needs of the power grid. For example, the particle swarm may find that increasing the discharge power of the energy storage during the evening peak period can effectively reduce the power purchase cost of the power grid. Real-time monitoring of the operating status of the energy storage group and the changes in the power flow of the power grid is the key to ensuring the stability of the power grid. Through continuous tracking, problems can be discovered in time and the control strategy can be adjusted. For example, if it is monitored that the voltage in a certain area of ​​the power grid begins to drop due to a sudden increase in load, the energy storage system can quickly adjust the strategy and increase the power supplied to the area to stabilize the voltage. If the control strategy may cause secondary disturbances in the power grid, such as excessive power regulation causing frequency fluctuations, the strategy adjustment mechanism needs to be triggered. The control strategy of the energy storage group can be dynamically updated through rolling optimization. This method takes into account the latest information on the power grid and energy storage status, generates a new optimal control strategy and issues it for execution. Such a dynamic adjustment mechanism ensures that the operation of the energy storage system is always consistent with the safety and stability requirements of the power grid. It is also necessary to continuously track the health status and life changes of the energy storage unit. For example, if the health indicator of a certain energy storage unit shows that its efficiency has dropped below a preset threshold, the system will automatically adjust the capacity and power limit of the unit participating in the grid regulation. This adjustment avoids overuse of the electrical energy storage device, thereby extending the service life of the device and maintaining the overall performance of the system.

[0031] Step S106, in accordance with the requirements for safe and stable operation of the power grid, the execution effect of the energy storage collaborative optimization control strategy is evaluated in real time. Combined with the dynamic changes of key indicators of grid frequency and voltage, the control strategy is optimized online using an adaptive weight adjustment mechanism. This mechanism adjusts the weight coefficients of different objective functions based on the dynamic evaluation results of the fifth step, continuously improves the effect of energy storage collaborative optimization control, and ensures safe and stable operation of the power grid.

[0032] The real-time operation data of the power grid, including key indicators such as power grid frequency and voltage, are obtained, and whether the power grid is in a safe and stable state is determined according to the preset threshold. If the operation state of the power grid is unstable, the energy storage collaborative optimization control strategy is triggered. A multi-objective optimization model is constructed based on the real-time operation data of the power grid and the parameters of the energy storage device. The objective functions include minimizing the frequency deviation of the power grid, minimizing the voltage deviation, etc., and the constraints include the capacity limit and power limit of the energy storage device. The particle swarm optimization algorithm is used to solve the multi-objective optimization model, obtain the optimal solution of the energy storage collaborative optimization control strategy, and send the control command to the energy storage device. The particle swarm optimization algorithm is suitable for solving multi-objective, nonlinear, and non-convex optimization problems. After the energy storage device executes the control command, feedback data is obtained from the device through the communication protocol, including actual operating power, energy storage status, etc. The fuzzy comprehensive evaluation method is used to calculate the control effect score, which can quantitatively evaluate the comprehensive effect of multiple indicators. If the control effect score is lower than the preset threshold, the adaptive weight adjustment mechanism is triggered. According to the difference between the score and the threshold, the gradient descent method is used to adjust the weight coefficients of different objective functions in the multi-objective optimization model to achieve a better control effect. The adjusted multi-objective optimization model is used as the new optimization target, and step 3 is re-executed until the control effect score reaches the preset requirements. According to the real-time operation data of the power grid and the control effect evaluation results, when the score is higher than the threshold for multiple consecutive times, the control strategy is dynamically updated. The reinforcement learning algorithm is used to use historical data as training samples to learn the optimal control strategy, realize online optimization of the strategy, and continuously ensure the safe and stable operation of the power grid.

[0033] Exemplarily, real-time grid operation data, such as frequency, voltage amplitude and phase angle, line flow, etc., are the basis for grid stability control. These data can be collected in real time through sensors and intelligent electronic devices (IDEs). For example, at a certain moment, the grid frequency is 49Hz, which is lower than the normal value of 50Hz, and the voltage in area A is 98pu, which is lower than the normal value of 0pu, indicating that the grid may be in an unstable state. Preset thresholds, such as frequency deviation ±2Hz and voltage deviation ±05pu, are used to determine whether the grid is safe and stable. When the grid operation state is unstable, for example, the frequency is lower than 48Hz or the voltage is lower than 95pu, it is necessary to trigger the electric energy storage collaborative optimization control strategy to maintain grid stability. At this time, a multi-objective optimization model needs to be constructed. The objective function may include minimizing the grid frequency deviation, minimizing the voltage deviation, and minimizing the energy storage loss. Constraints include capacity limit, charge and discharge power limit, charge and discharge rate limit, etc. of each energy storage unit. For example, the capacity of a certain energy storage unit is 1MWh, the maximum charge and discharge power is 5MW, and the maximum charge and discharge rate is 2MW / min. Particle swarm optimization is a heuristic algorithm that can be used to solve multi-objective optimization problems. Each "particle" represents a possible control strategy, such as the charge and discharge power of each energy storage unit. By iteratively updating the position and velocity of the particles, the algorithm can search for the optimal solution in the feasible domain. For example, the algorithm may find a solution that causes energy storage unit A to discharge 3MW and energy storage unit B to discharge 2MW, thereby restoring the grid frequency to 50Hz and increasing the voltage in area A to 0pu. Control instructions, such as charge and discharge power set values, are sent to each energy storage unit through the communication system. After the energy storage unit executes the instruction, it will feed back the actual operating data, such as the actual charge and discharge power, energy storage status, etc. to the control center. The control effect score is calculated using the fuzzy comprehensive evaluation method. For example, the weights of the three indicators of frequency deviation, voltage deviation and energy storage loss are set to 5, 3 and 2 respectively. If the frequency deviation is 1Hz, the voltage deviation is 02pu, and the energy storage loss is 01MWh, the score of each indicator can be calculated according to the membership function, and then the weighted average is used to obtain the final control effect score. If the control effect score is lower than the preset threshold, such as 8, the adaptive weight adjustment mechanism is triggered. For example, if the score is 7, the weight coefficient of the objective function can be adjusted according to the gradient descent method. For example, the weight of the frequency deviation can be increased to 6, the weight of the voltage deviation can be reduced to 2, and the weight of the energy storage loss can remain unchanged. Then, the multi-objective optimization model is re-solved using the new weight coefficients. When the control effect score is higher than the preset threshold for multiple consecutive times, such as higher than 9 for 5 consecutive times, the dynamic update mechanism of the control strategy is triggered. The reinforcement learning algorithm can learn and update the control strategy based on historical data and the current power grid status. For example, based on the power grid load changes and energy storage charging and discharging data in the past period of time, the power grid operation status in the future period can be predicted and the corresponding control strategy can be formulated.In this way, the coordinated optimization control of electric energy storage can continuously ensure the safe and stable operation of the power grid. The health status and life of the energy storage unit are also important considerations. If the health status of a certain energy storage unit is lower than the preset threshold, its charge and discharge power limit needs to be reduced to avoid overuse and extend its service life.

[0034] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations. In addition, the various different embodiments of the present invention can also be combined arbitrarily, as long as they do not violate the concept of the present invention, they should also be regarded as the contents disclosed by the present invention.

Claims

1. A method for processing power grid disturbance signals and optimizing electric energy storage control, characterized in that: include: Adaptive Kalman filter algorithm is used to filter the second-level disturbance signal in the real-time operation data of the power grid to improve the signal-to-noise ratio; The wavelet analysis method is used to extract the time-frequency domain features of the filtered disturbance signal, and the type and severity of the disturbance signal are determined in combination with the preset disturbance pattern library; According to the type and characteristics of the disturbance signal, an electric energy storage unit with complementary dynamic characteristics is selected from the electric energy storage group, and an online identification model of the dynamic characteristics of the electric energy storage is established; The robust optimization theory is used to construct an energy storage collaborative optimization control model, and the control strategy of the energy storage group is adjusted in real time to ensure that the charging and discharging behavior of the energy storage group will not cause secondary disturbances to the power grid; The execution effect of the coordinated optimization control strategy of electric energy storage is evaluated in real time, and the control strategy is optimized online using an adaptive weight adjustment mechanism according to the requirements of safe and stable operation of the power grid.

2. The method according to claim 1, characterized in that The method of using an adaptive Kalman filter algorithm to filter the second-level disturbance signal in the real-time operation data of the power grid includes: The real-time operation data of the power grid is obtained, and the initial measurement noise covariance matrix and the process noise covariance matrix are estimated using the least squares method. The difference between the estimated value and the measured value of the disturbance signal is calculated in real time to obtain the residual sequence. The values ​​of the measurement noise covariance matrix and the process noise covariance matrix are adjusted in real time according to the mean and variance of the residual sequence.

3. The method according to claim 1, characterized in that The method of extracting time-frequency domain features of the filtered disturbance signal using a wavelet analysis method comprises: The filtered disturbance signal is subjected to wavelet transform to obtain the energy distribution matrix of the signal in the time-frequency domain, extract the time-frequency region with relatively concentrated energy, determine the main frequency components in the region, and construct a multi-dimensional feature vector.

4. The method according to claim 1, characterized in that The step of selecting an electric energy storage unit with complementary dynamic characteristics from the electric energy storage group according to the type and characteristics of the disturbance signal comprises: According to the response speed, duration and power capacity of each unit in the electric energy storage group, particle swarm optimization or genetic algorithm is used to select the optimal combination from the electric energy storage group to achieve complementary dynamic characteristics.

5. The method according to claim 1, characterized in that The robust optimization theory is used to construct an electric energy storage collaborative optimization control model, including: By analyzing the historical operation data of the electric energy storage system, a mathematical description considering uncertainty is established. The established electric energy storage dynamic model with uncertainty is used as a constraint condition, and the NSGA-II multi-objective optimization algorithm is used to solve the multi-objective optimization problem considering uncertainty, and the Pareto optimal solution set of the collaborative optimization control of the electric energy storage system is obtained.

6. The method according to claim 1, characterized in that The real-time evaluation of the execution effect of the electric energy storage collaborative optimization control strategy includes: Obtain real-time grid operation data, including grid frequency and voltage, determine whether the grid is in a safe and stable state based on preset thresholds, and use fuzzy comprehensive evaluation methods to calculate the control effect score. If the control effect score is lower than the preset threshold, the adaptive weight adjustment mechanism is triggered.

7. The method according to claim 2, characterized in that The real-time adjustment of the values ​​of the measurement noise covariance matrix and the process noise covariance matrix according to the mean and variance of the residual sequence includes: The sliding window method is used to calculate the mean and variance of the residual sequence. If the residual mean exceeds the preset threshold or the residual variance exceeds the preset threshold, it is judged that the noise environment has changed; When the noise environment changes, the maximum likelihood estimation method is used to re-estimate the values ​​of the measurement noise covariance matrix R and the process noise covariance matrix Q based on the residual mean and variance; Use the updated R and Q as new input parameters to update the state equation and observation equation of the Kalman filter algorithm; Repeat the process of residual calculation, noise environment judgment and parameter adaptive adjustment until the residual mean and variance in N consecutive windows meet the preset convergence conditions, that is, the residual mean is less than ε1 and the residual variance is less than ε2, then the filtering process is considered to converge, and the estimated value of the disturbance signal after removing the influence of measurement noise is obtained; Calculate the signal-to-noise ratio of the disturbance signal before and after filtering, and evaluate the effect of the adaptive Kalman filter algorithm on improving the signal-to-noise ratio; If the improvement in the signal-to-noise ratio does not meet expectations, the values ​​of ε1 and ε2 in the convergence condition and the size of the sliding window are adjusted according to the improvement in the signal-to-noise ratio, and the parameter settings of the filtering algorithm are further adjusted.

8. The method according to claim 6, characterized in that The adaptive weight adjustment mechanism comprises: Through rolling optimization, the control strategy of the energy storage group is dynamically updated, and a new optimal control strategy is generated and issued for execution to ensure that it is compatible with the safety and stability constraints of the power grid; Continuously track the health status and life span changes of the energy storage unit. When the health index of the energy storage unit is lower than the preset threshold, adjust its capacity and power limit for participating in grid regulation.

9. The method according to claim 1, characterized in that The control strategy for adjusting the electric energy storage group in real time includes: Continuously monitor the operating status of the electric energy storage system and the disturbance signal of the power grid. When a significant change in the operating status or disturbance signal is detected, the update of the dynamic characteristic model and uncertainty description of the electric energy storage is triggered. The model is adaptively adjusted online through incremental learning, and the multi-objective optimization solution is re-performed to generate updated optimal control instructions, thereby realizing the adaptive update of the collaborative optimization control of the electric energy storage.

10. The method according to claim 1, characterized in that The online identification model for establishing the dynamic characteristics of electric energy storage comprises: Obtain second-level disturbance information in the power grid, and determine the power support and response time requirements of the energy storage unit based on the type and characteristics of the disturbance; According to the dynamic characteristic parameters of each unit in the energy storage group, including response speed, duration and power capacity, combined with the power support and response time requirements determined in the first step, a dynamic characteristic model of the energy storage group is established; Real-time collection of operating data of the electric energy storage unit, including power output and energy status, as input to the online identification model; Through the recursive least squares method or Kalman filter online identification algorithm, the dynamic characteristic model of the electric energy storage is corrected in real time to obtain an accurate description of the dynamic characteristics.

Images