Power grid frequency modulation cooperative control method giving consideration to tie line section

Abstract

The invention provides a power grid frequency modulation cooperative control method considering a tie line section, and relates to the field of power system automatic scheduling, and the method comprises the steps: obtaining a load characteristic parameter, a transmission capacity parameter and a stability margin parameter of a section spanned by a tie line; solving the multi-objective optimization model by adopting an improved particle swarm optimization algorithm to obtain an optimal tie line electric power distribution curve meeting various constraint conditions; discretizing the optimal tie line electric power distribution curve, and converting the discretized optimal tie line electric power distribution curve into an active power instruction value of a tie line control system; and the tie line control system adjusts the electric power of the tie line in real time according to the active power instruction value. According to the invention, the reasonable distribution of the electric power of the tie line is realized, the frequency modulation demand and transmission capacity limitation of each section are balanced, and the safety and stability of power grid operation are improved. Through a differential scheduling strategy and a real-time optimization mechanism, load fluctuation can be effectively coped with, and safe and stable operation of a power grid is guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of automatic dispatching of electric power systems, and in particular to a coordinated control method for frequency regulation of power grids taking into account tie line sections. Background Art

[0002] In the coordinated control of power grid frequency regulation, the tie line is an important channel connecting different regional power grids, and the optimal configuration of its electric power is crucial to achieve effective frequency regulation of the entire network. However, due to the significant differences in the characteristics of the sections crossed by the tie line, how to take into account the needs of different sections and achieve reasonable allocation of the tie line power is a technical problem that needs to be solved urgently.

[0003] Specifically, the sections spanned by the tie lines have significant differences in load characteristics, transmission capacity, and stability margin. On the one hand, the different load characteristics lead to different demands for frequency regulation capabilities in each section. Sections with drastic load fluctuations require stronger frequency regulation capabilities, while sections with relatively stable loads have lower requirements for frequency regulation capabilities. On the other hand, due to the constraints of the overload capacity and static stability limit of the transmission line, the transmission capacity of different sections varies significantly. As an important indicator for measuring the safe and stable operation capability of the power grid, the stability margin is also unbalanced between different sections.

[0004] The differences in the above-mentioned section characteristics make the optimal configuration of the interconnection line power face multiple contradictions and challenges. How to meet the frequency regulation needs of each section while taking into account the constraints of transmission capacity and stability margin to achieve the reasonable distribution of interconnection line power among different sections is a complex combinatorial optimization problem. This requires in-depth analysis of section characteristics, comprehensive consideration of multiple objectives such as frequency regulation capability, transmission capacity and stability margin, establishment of a multi-objective optimization model, and design of an efficient solution algorithm to obtain the optimal solution that meets multiple constraints, achieve the reasonable configuration of interconnection line power, and thus ensure the safe and stable operation of the power grid. Summary of the invention

[0005] The present invention provides a coordinated control of power grid frequency regulation taking into account the tie line section, which mainly includes:

[0006] Obtain the load characteristic parameters, transmission capacity parameters and stability margin parameters of the section spanned by the tie line; establish a multi-objective optimization model based on the load characteristic parameters, transmission capacity parameters and stability margin parameters, wherein the objective function of the multi-objective optimization model is to minimize the mean square error of the tie line electric power, and the constraints include load prediction curve constraints, transmission capacity constraints, stability margin constraints, frequency regulation capability constraints and overload capability constraints; use an improved particle swarm optimization algorithm to solve the multi-objective optimization model to obtain an optimal tie line electric power distribution curve that meets various constraints; discretize the optimal tie line electric power distribution curve and convert it into an active power command value of a tie line control system; the tie line control system adjusts the electric power of the tie line in real time according to the active power command value to achieve optimal control of the tie line flow.

[0007] Furthermore, the step of obtaining the load characteristic parameters of the section spanned by the tie line includes: establishing a load prediction model based on the historical load data of the tie line section using a long short-term memory neural network to obtain a load prediction curve for a period of time in the future; determining the maximum transmission power threshold of the tie line based on the physical parameters and thermal stability limits of the tie line, and using the threshold as the transmission capacity constraint condition of the multi-objective optimization model.

[0008] Furthermore, based on the historical load data of the interconnection line section, the method includes: processing the data using preprocessing techniques such as data cleaning and normalization to obtain a standardized historical load data set; constructing a feature set based on the preprocessed historical load data, and selecting load-related features as model input; adopting an LSTM neural network algorithm, setting network structure parameters and training parameters, and optimizing the feature data training to obtain a load prediction model; in an optimization flow model with the goal of minimizing power generation costs, using the maximum transmission power constraint as the upper limit constraint condition of the interconnection line flow, and combining the power balance constraint and the upper and lower limit constraints of the generator output to construct an optimization model.

[0009] Furthermore, the step of using the improved particle swarm optimization algorithm to solve the multi-objective optimization model includes: introducing adaptive inertia weights and learning factors, adopting a non-dominated sorting strategy, and dynamically balancing the frequency regulation requirements and transmission capacity limitations of each section during the search process by designing a fitness function and an update strategy, so as to obtain the optimal interconnection line power distribution curve that meets multiple constraints.

[0010] Furthermore, according to the physical parameters of the line, a power flow calculation method is used to obtain the transmission capacity limit threshold of the line, obtain historical load data, and predict the load demand in the future through a time series prediction algorithm such as an ARIMA model; according to the transmission capacity limit threshold and the load forecast result, a preset frequency regulation capacity estimation formula is used to calculate the required frequency regulation capacity demand; a multi-objective optimization model with power generation cost and spare capacity as objective functions and with unit output upper and lower limits and ramp rate limits as constraints is established, and a mixed integer programming solving tool is used to solve the model to obtain the optimal unit output and spare capacity configuration plan; the grid frequency measured in real time by the grid frequency measurement device is obtained, and if the frequency deviation exceeds the dead zone threshold specified in the preset grid frequency regulation standard, a new round of frequency regulation control process is triggered.

[0011] Furthermore, after obtaining the optimal interconnection line power distribution curve, it also includes: according to the electrical parameters of the interconnection line, using the continuous power angle stability margin and transient stability margin evaluation method to calculate the stability margin index of the optimal interconnection line power distribution curve; judging whether the stability margin index meets the stability requirements, if not, taking the stability margin index as a new constraint condition and returning to the multi-objective optimization model for re-solving.

[0012] Furthermore, the step of adjusting the electric power of the interconnection line in real time includes: obtaining real-time load data and transmission power data of each section of the interconnection line, and determining whether the deviation exceeds a preset threshold; if it exceeds the preset threshold, triggering the multi-objective optimization model to recalculate, generate a new active power command value, and adjust the electric power of the interconnection line until the deviation value is less than the preset threshold.

[0013] Furthermore, after obtaining the optimal interconnection line power distribution curve, it also includes: formulating differentiated interconnection line power dispatching strategies based on the differences in section characteristics; increasing the power transmitted by the interconnection line to sections with severe load fluctuations; and reducing the transmission power to sections with relatively stable loads, thereby achieving reasonable allocation of frequency regulation resources.

[0014] Furthermore, the step of establishing a multi-objective optimization model includes: taking the minimum mean square error of the interconnection line electric power as the objective function, and taking the load prediction curve, transmission capacity constraint, stability margin constraint, frequency regulation capability constraint and overload capability constraint as constraints to construct a multi-objective optimization model; the load prediction curve is obtained based on the historical data and prediction algorithm of the load of each section, the transmission capacity constraint is calculated based on the physical parameters of the interconnection line, the stability margin constraint is determined based on the electrical parameters of the interconnection line and the margin evaluation method, the frequency regulation capability constraint is determined based on the frequency regulation capacity and load characteristics of the power grid, and the overload capability constraint is determined based on the line conductor parameters and environmental conditions.

[0015] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0016] The present invention discloses a method for coordinated control of power grid frequency regulation taking into account interconnection line sections. The method establishes a multi-objective optimization model that takes into account frequency regulation requirements, transmission capacity, and stability margin, and adopts an improved particle swarm algorithm for solving the problem. By real-time monitoring of the load level of each section, the interconnection line power distribution is dynamically adjusted. Scenario analysis and risk assessment are introduced to improve the risk resistance of the solution. A mapping relationship between deviations and influencing factors is established using a support vector machine or a neural network algorithm to provide a basis for dynamic optimization. The present invention realizes the reasonable distribution of interconnection line power, balances the frequency regulation requirements of each section and the transmission capacity limitations, and improves the safety and stability of power grid operation. Through differentiated scheduling strategies and real-time optimization mechanisms, the present invention can effectively cope with load fluctuations and ensure safe and stable operation of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The present invention is a flow chart of a method for coordinated control of power grid frequency regulation taking tie line sections into consideration. DETAILED DESCRIPTION

[0018] In order to further understand the content of the present invention, the present invention is described in detail in conjunction with the accompanying drawings and embodiments. The present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It is understood that the specific embodiments described herein are only used to explain the relevant inventions, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the invention are shown in the accompanying drawings.

[0019] like Figure 1 In this embodiment, a method for coordinated control of power grid frequency regulation taking tie line sections into consideration may specifically include:

[0020] Step S101, a multi-objective optimization model is established based on the load characteristics, transmission capacity and stability margin parameters of the section spanned by the tie line. The model comprehensively considers the frequency regulation capability requirements, the overload capacity of the transmission line and the static stability limit constraints, with the optimization goal of achieving the reasonable allocation of the tie line power.

[0021] According to the historical load data of the tie line section, a load prediction model is established by using a long short-term memory neural network (LSTM) to predict the load changes in the future and obtain the load prediction curve. According to the physical parameters and thermal stability limit of the tie line, the maximum transmission power threshold of the tie line is determined, and this threshold is used as one of the constraints. According to the electrical parameters of the tie line, the stability margin index of the tie line is calculated by using the continuous power angle stability margin and transient stability margin evaluation methods to determine whether the tie line meets the stability requirements. The stability margin index is converted into a power constraint condition as one of the constraints of the optimization model. According to the historical data of the frequency regulation capacity of the power grid, a frequency regulation capacity prediction model is established by using support vector machine regression (SVR) to predict the frequency regulation capacity demand in the future, and the prediction results are used as the power balance constraint condition of the optimization model. According to the conductor parameters and environmental conditions of the transmission line, the thermal stability limit of the transmission line is determined as the maximum load threshold, and this threshold is used as another constraint condition of the optimization model. Taking the minimum mean square error of the tie line power as the objective function, and the load prediction curve, transmission capacity constraint, stability margin constraint, frequency regulation capacity constraint and overload capacity constraint as constraints, a multi-objective optimization model is constructed. The particle swarm optimization algorithm is used to solve the model and obtain the optimal tie-line power distribution curve that meets various constraints. The optimized tie-line power distribution curve is discretized and converted into active power command values ​​for each time period, which are then sent to the tie-line control system. The control system adjusts the tie-line power in real time according to the power command value to achieve optimal control of the tie-line power flow.

[0022] Specifically, tie-line load prediction is the basis of power flow optimization control. Using the existing tie-line section historical load data, a long short-term memory neural network (LSTM) can be used to establish a load prediction model. LSTM neural networks are good at processing time series data and can capture long-term and short-term dependencies in load data. For example, the power load in a region usually has obvious daily and weekly periodicities. LSTM can learn these patterns and predict the load changes in the future, and finally obtain a smooth load prediction curve. For example, the prediction results show that in the next 24 hours, the load will reach a peak in the evening and drop to a trough in the middle of the night. Determining the maximum transmission power threshold of the tie line requires considering the physical parameters and thermal stability limit of the tie line. Physical parameters include conductor cross-sectional area, length, material, etc. The thermal stability limit refers to the maximum power that the tie line can safely transmit without exceeding the allowable temperature. For example, a 500kV tie line, based on its conductor parameters and ambient temperature, has a calculated thermal stability limit of 2000MW. Exceeding this value, the conductor may fail due to overheating. This threshold will be used as one of the constraints of the subsequent optimization model. To evaluate the stability of the tie line, it is necessary to consider the continuous power angle stability margin and the transient stability margin. The power angle stability margin refers to the ability of the system to restore stability after a small disturbance, and the transient stability margin refers to the ability of the system to remain stable after a large disturbance. These stability margin indicators can be calculated using professional power system analysis software based on the electrical parameters of the tie line, such as line impedance and node voltage. For example, the calculation results show that the continuous power angle stability margin of the current tie line is 20% and the transient stability margin is 15%. These two indicators need to meet certain safety standards, for example, the power angle stability margin is not less than 10%, and the transient stability margin is not less than 8%. These indicators are converted into power constraints. For example, in order to ensure that the transient stability margin is not less than 15%, the transmission power of the tie line cannot exceed 1800MW. To predict the demand for frequency regulation capacity in the future, support vector machine regression (SVR) can be used to establish a frequency regulation capacity prediction model. SVR is a powerful machine learning algorithm that can handle nonlinear relationships. Frequency regulation capacity refers to the ability of the power system to regulate frequency, which is related to factors such as the system's power generation capacity and load characteristics. For example, the forecast results show that in the next 24 hours, due to the arrival of the peak electricity consumption, the system's frequency regulation capacity demand will increase from the current 500MW to 600MW. This forecast result will be used as a power balance constraint for the optimization model to ensure that the operation of the tie line will not affect the system's frequency regulation capacity. The thermal stability limit of the transmission line determines its maximum load threshold. Based on the conductor parameters of the transmission line, such as conductor type, cross-sectional area, etc., and environmental conditions, such as ambient temperature, wind speed, etc., the maximum load that the transmission line can safely carry can be determined.For example, according to calculations, the thermal stability limit of a transmission line in high temperature weather in summer is 1500MW. This value will be used as another constraint of the optimization model to prevent line overload. The goal of building a multi-objective optimization model is to minimize the mean square error of the tie line power. The smaller the mean square error, the more stable the power output of the tie line, which is conducive to the stable operation of the system. The constraints include load forecast curve, transmission capacity constraint, stability margin constraint, frequency regulation capability constraint and overload capacity constraint. For example, the objective function is to minimize the mean square error of the tie line power output in the next 24 hours. The constraints include: the tie line power output must meet the load forecast curve, the transmission power cannot exceed 2000MW, the stability margin requirements must be met, the frequency regulation capability of the system must not be affected, and the thermal stability limit of the transmission line must not be exceeded. The particle swarm optimization algorithm is used to solve the model to obtain the optimal tie line power distribution curve that meets various constraints. The particle swarm optimization algorithm is an intelligent optimization algorithm that simulates the foraging behavior of bird flocks and has strong global search capabilities. Through iterative calculations, the algorithm can find an optimal solution that minimizes the objective function and meets all constraints. For example, the optimization result is a smooth power distribution curve, which minimizes the fluctuation of the tie line power output under the premise of satisfying all constraints. Finally, the optimized tie line power distribution curve is discretized and converted into active power command values ​​for each time period, which are sent to the tie line control system. For example, the power distribution curve for the next 24 hours is discretized into 96 time points, each of which corresponds to a 15-minute power command value. The control system adjusts the electric power of the tie line in real time according to these command values, realizes the optimal control of the tie line flow, and thus improves the operating efficiency and safety of the system.

[0023] Step S102, based on the historical load data of the tie line section, a load prediction model is established using a long short-term memory neural network to obtain a load prediction curve for a period of time in the future, determine the maximum transmission power threshold of the tie line, and use the threshold as one of the constraints of the optimization model.

[0024] According to the historical load data of the tie line section, the data is processed by preprocessing techniques such as data cleaning and normalization to obtain a standardized historical load data set. For the preprocessed historical load data, a feature set is constructed, and the features related to the load are selected as the model input. The long short-term memory (LSTM) neural network algorithm is used to construct the load prediction model. By setting appropriate network structure parameters and training parameters, the model is trained and optimized using the prepared feature data to obtain a load prediction model with good performance. The trained LSTM load prediction model is used to predict the load in the future period, and the load prediction curve and the maximum load value in the period are obtained. The maximum value is used as the preliminary maximum transmission power threshold of the tie line. According to the physical parameters of the tie line, such as the conductor cross-sectional area, material, permitted current carrying capacity, etc., and considering a certain safety margin, the preliminary determined maximum transmission power threshold is checked and necessary corrections are made to obtain the final maximum transmission power constraint value of the tie line. In the optimization power flow model with the goal of minimizing the power generation cost, the maximum transmission power constraint is used as the upper limit constraint condition of the tie line power flow, and the power balance constraint of the power system and the upper and lower limit constraints of the generator output are considered to construct a complete optimization model. The particle swarm optimization algorithm is used to solve the optimization model and obtain the optimal power flow allocation scheme under the condition of satisfying the maximum transmission power constraint of the tie line to guide the actual system operation.

[0025] Specifically, data cleaning and normalization are important prerequisites for ensuring the accuracy of load prediction models. Historical load data in power systems often contain outliers and noise, such as spikes or missing values ​​caused by measurement errors or equipment failures. Data cleaning techniques can identify and process these abnormal data, such as using mean filtering or median filtering to remove noise and using interpolation to fill missing values. For example, an abnormally high value appears in the historical load data of a certain day, which may be due to a surge in power consumption caused by the restoration of power supply after a short power outage. Through data cleaning, it can be identified and corrected to avoid interference with model training. Data normalization can convert data of different dimensions into the same numerical range, such as scaling all data to between 0 and 1, to avoid some features from having too much impact on model training due to excessive values. For example, temperature and humidity are both important factors affecting power load, but the numerical range of temperature is much larger than that of humidity. If normalization is not performed, the temperature feature will dominate the model training, while the influence of humidity will be weakened. Through data cleaning and normalization, a standardized historical load data set can be obtained, laying the foundation for subsequent model training. To construct a feature set, you need to select appropriate features based on specific business scenarios. There are many factors that affect power load, such as weather, date, time, holidays, economic activities, etc. The correlation between different features and load can be analyzed based on historical data, and features with higher correlation can be selected as model inputs. For example, when the temperature is high in summer, the air conditioning load will increase significantly, so meteorological factors such as temperature, humidity, and sunshine time can be selected as features. In addition, date type (weekdays, weekends, holidays) and time period (daytime, nighttime) will also affect the load, and these factors can be encoded as feature inputs to the model. A typical feature set may include date type, time period, temperature, humidity, sunshine time, and load data from the previous day or week. Long short-term memory (LSTM) neural networks are suitable for processing time series data, such as power load data. LSTM can capture long-term and short-term dependencies in the data. For example, power load usually has obvious daily and weekly periodicities. LSTM can learn these patterns and predict load changes in the future. For example, the power load in a region usually reaches a peak in the evening and drops to a trough in the middle of the night. LSTM can learn this daily periodic variation and take this periodicity into account when predicting future loads. Suppose an LSTM model is trained using historical load data from the past year. The input features include date type, time period, temperature, humidity, and load data from the previous day. The output is the load forecast for the next 24 hours. By adjusting the structural parameters of the LSTM network, such as the number of hidden layers and neurons, and training parameters such as the learning rate and batch size, the model's prediction performance can be optimized so that it can accurately predict future load change trends.The trained LSTM load prediction model can be used to predict the load curve in the future, such as the load curve in the next week or the next month. The maximum load value in the future can be obtained through the load prediction curve, and the maximum value is used as the preliminarily determined maximum transmission power threshold of the tie line. For example, if the prediction result shows that the maximum load value in the next week is 1800MW, the preliminarily determined maximum transmission power threshold of the tie line can be 1800MW. In order to ensure the safe and stable operation of the tie line, it is necessary to check and correct the preliminarily determined maximum transmission power threshold according to the physical parameters of the tie line. The physical parameters of the tie line include the conductor cross-sectional area, material, and permitted current carrying capacity. For example, the permitted current carrying capacity of a 500kV tie line is 2000MW. Even if the maximum load prediction value is 1800MW, the maximum transmission power threshold needs to be limited to within 2000MW to ensure that the tie line will not be overloaded. In addition, a certain safety margin needs to be considered, such as setting the maximum transmission power threshold to 90% of the permitted current carrying capacity, that is, 1800MW. The final maximum transmission power constraint value is 1800MW, which not only meets the load demand but also ensures the safe and stable operation of the tie line. In the optimization power flow model, with the goal of minimizing the power generation cost, the maximum transmission power constraint of the tie line is used as the upper limit constraint of the power flow, and is solved in combination with other constraints of the power system. The constraints of the power system include power balance constraints, upper and lower limit constraints of generator output, and upper and lower limit constraints of node voltage. The power balance constraint requires that the total power generation of the system is equal to the total load plus network loss. The upper and lower limit constraints of generator output limit the output range of each generator. The upper and lower limit constraints of node voltage ensure the voltage stability of the power system. For example, suppose a power system contains two generators, a tie line and a load, and the goal is to minimize the total power generation cost of the two generators. The constraints include: the sum of the output of the two generators is equal to the load plus the network loss, the power flow of the tie line cannot exceed 1800MW, the output of the two generators is within their upper and lower limits, and the voltage of the load node is within the allowable range. The particle swarm optimization algorithm is used to solve the optimization model, and the optimal power flow distribution scheme that meets all constraints can be obtained. For example, the output of generator A is 800MW, the output of generator B is 1000MW, the power flow of the tie line is 800MW, the load is 1800MW, and the network loss is 0. This scheme minimizes the power generation cost while meeting the maximum transmission power constraint of the tie line, and ensures the safe and stable operation of the power system.

[0026] Step S103, using an improved particle swarm optimization algorithm to solve the established multi-objective optimization model. The improvements include introducing adaptive inertia weights and learning factors, and adopting a non-dominated sorting strategy. By designing a fitness function and an update strategy, the frequency modulation requirements and transmission capacity constraints of each section are dynamically balanced during the search process, and an optimal configuration scheme for interconnection line power that meets multiple constraints is obtained.

[0027] According to the multi-objective optimization model, the fitness function of the particle swarm optimization algorithm is established, and the frequency regulation demand and transmission capacity limit of each section are used as constraints. Among them, the frequency regulation demand is obtained according to the load prediction results of each section, and the transmission capacity limit is calculated according to the physical parameters of the line. Initialize the particle swarm and randomly generate a set of interconnection line power configuration schemes that meet the frequency regulation demand and transmission capacity constraints as the initial particle positions. The fitness value of each particle is calculated by the weighted sum method, and the particles are sorted according to the non-dominated sorting strategy, and the individual optimal position and global optimal position of the particles are updated. Non-dominated sorting divides particles into different non-dominated levels according to the fitness values ​​of particles on each objective function, and gives priority to particles with lower levels. According to the adaptive inertia weight and learning factor, the speed of the particle is updated using the speed update formula, and then the position of the particle is updated according to the position update formula to generate a new interconnection line power configuration scheme. The adaptive inertia weight decreases linearly according to the current number of iterations, and the learning factor adopts a fixed value. Determine whether the newly generated particles meet the frequency regulation demand and transmission capacity constraints. If not, regenerate them until all particles meet the constraints. Repeat the steps of particle fitness calculation, particle position and velocity update, and new particle generation until the maximum number of iterations is reached or the termination condition is met. The termination condition is set according to the convergence of the fitness value. When the change in the optimal fitness value of multiple consecutive iterations is less than the set threshold, the algorithm is considered to have converged. The particle with the best fitness value is selected from the particle swarm at the time of convergence, and the corresponding interconnection line power configuration scheme is used as the optimal scheme and applied to the actual power system. According to the optimal configuration scheme, the active power of each section is adjusted in real time to realize the interconnection line power control under multi-objective optimization, and ensure the frequency stability and power balance of the system.

[0028] Specifically, the multi-objective optimization control of the interconnection line power distribution aims to achieve the rationality and economy of power distribution under the premise of meeting the safe and stable operation of the system. Particle swarm optimization algorithm is an effective method for solving multi-objective optimization problems. First, it is necessary to establish the fitness function of the particle swarm optimization algorithm. The fitness function is used to evaluate the pros and cons of each particle (representing a interconnection line power distribution scheme). It is usually formed by the weighted summation of multiple objective functions. For example, if the goal is to minimize the interconnection line power fluctuation and transmission loss at the same time, the fitness function can be expressed as: fitness = w1*power fluctuation + w2*transmission loss, where w1 and w2 are weight coefficients used to balance the importance of the two objectives. Assume that in a certain period of time, the power fluctuation of scheme A is 100MW and the transmission loss is 5MW; the power fluctuation of scheme B is 150MW and the transmission loss is 2MW. If w1=0.8 and w2=0.2 are set, the fitness value of Scheme A is 0.8*100+0.2*5=81, and the fitness value of Scheme B is 0.8*150+0.2*2=120.4. Since Scheme A has a smaller fitness value, Scheme A is considered to be better. Frequency regulation demand and transmission capacity limit are important constraints. Frequency regulation demand refers to the frequency regulation capability required by the system in a certain period of time, which can be obtained based on the load forecast results of each section. For example, if the forecast shows that the load of a section will increase by 100MW in the next hour, the section needs at least 100MW of frequency regulation capability to maintain frequency stability. The transmission capacity limit is determined by the physical parameters of the line, such as the conductor cross-sectional area, length, material, etc. The transmission capacity of a 500kV line may be 2000MW, and the line may be overloaded if it exceeds this value. When initializing the particle swarm, a set of initial particles that meet these constraints need to be generated. For example, 100 interconnection line power configuration schemes are randomly generated, each of which must meet the frequency regulation requirements of each section, and the transmission power of any line cannot exceed its capacity limit. The non-dominated sorting strategy is used to sort particles. Non-dominated sorting divides particles into different levels according to their performance on various objective functions. If particle A is better than particle B on all objective functions, then A is said to dominate B. Non-dominated sorting gives priority to particles with lower levels because these particles perform better on at least one objective function and no other particles are better than them on all objective functions. For example, suppose there are two objective functions: minimizing power fluctuations and minimizing transmission losses. Particle A has a power fluctuation of 100MW and a transmission loss of 5MW; particle B has a power fluctuation of 150MW and a transmission loss of 2MW; particle C has a power fluctuation of 80MW and a transmission loss of 6MW. Then C dominates A, B does not dominate any particle, and A does not dominate any particle. Therefore, C is in the first level, and A and B are in the second level. The particle swarm optimization algorithm finds the optimal solution by iteratively updating the position and velocity of particles.The speed update formula contains parameters such as inertia weight, learning factor, individual optimal position, and global optimal position. The inertia weight controls the degree to which the particle inherits the previous speed, and usually decreases with the increase in the number of iterations. The learning factor controls the degree to which the particle learns from the individual optimal position and the global optimal position. For example, assuming that the current particle speed is 10MW, the inertia weight is 0.8, the learning factor is 0.5, the power corresponding to the individual optimal position is 150MW, and the power corresponding to the global optimal position is 180MW. Then the new speed of the particle can be calculated as: 0.8*10+0.5*(150-current power)+0.5*(180-current power). The position update formula updates the position of the particle according to its speed, that is, the new power configuration scheme. In each iteration, it is necessary to check whether the newly generated particles meet the constraints. If not, they need to be regenerated until they are satisfied. For example, if the newly generated scheme causes a line to be overloaded, the power allocation on the line needs to be adjusted until its transmission power is within its capacity limit. The iteration process continues until the maximum number of iterations is reached or the termination condition is met. The termination condition is usually set according to the convergence of the fitness value. For example, if the optimal fitness value changes by less than 0.1% for 10 consecutive iterations, the algorithm is considered to have converged. Finally, the particle with the best fitness value is selected from the particle swarm at the time of convergence, and its corresponding interconnection line power configuration scheme is the optimal scheme. This scheme will be used to adjust the active power of each section in real time, realize the interconnection line power control under multi-objective optimization, and ensure the frequency stability and power balance of the system. For example, the optimal scheme may recommend adjusting the power of a certain interconnection line from 1000MW to 1200MW to better meet the frequency regulation requirements of a certain section and reduce the transmission loss of the entire system.

[0029] Step S104, calculate the transmission capacity limit based on the line physical parameters, determine the frequency regulation demand in combination with the section load prediction result, obtain the optimal configuration plan through multi-objective optimization, and use this plan to adjust the active power in real time to ensure frequency stability and power balance.

[0030] According to the physical parameters of the line, the power flow calculation method is used to obtain the transmission capacity limit threshold of the line. The historical load data is obtained, and the load demand for a period of time in the future is predicted by time series prediction algorithms such as the ARIMA model. According to the transmission capacity limit threshold and the load prediction results, the predetermined frequency regulation capacity estimation formula is used to calculate the required frequency regulation capacity demand. A multi-objective optimization model is established, in which the objective function includes power generation cost and spare capacity, and the constraints include the upper and lower limits of unit output, ramp rate limit, etc. The model is solved by a mixed integer programming solver to obtain the optimal unit output and spare capacity configuration scheme. According to the output value of each unit in the optimal configuration scheme, the active power adjustment amount of each power station is calculated, and the active power adjustment instruction is formed and sent to the power station control system. According to the adjustment instruction, the control system adjusts the active power output of the power station in real time by adjusting the opening of the steam inlet valve of the steam turbine. The grid frequency is obtained in real time through the grid frequency measurement device. If the frequency deviation exceeds the dead zone threshold specified in the grid frequency regulation standard, a new round of frequency regulation control process is triggered. The control system automatically adjusts the active power of the power station according to the frequency deviation direction and the preset frequency modulation rate until the frequency returns to the dead zone. If the frequency still exceeds the threshold after frequency modulation, the load forecast value is corrected according to the frequency modulation result, and the load is re-forecasted in the second step and the frequency modulation capacity demand is calculated, and the next round of frequency modulation is performed until the frequency returns to stability.

[0031] Specifically, the transmission capacity limit threshold of the line is determined by the power flow calculation method, combined with the physical parameters of the line, such as the conductor cross-sectional area, length, material, etc. It can be understood as the maximum power that the line can withstand when it is operating safely and stably. For example, for a 500kV line, based on its physical parameters and power flow calculation results, its transmission capacity limit threshold is determined to be 2000MW. If this value is exceeded, the line may be at risk of overheating or even tripping. Load forecasting is an important basis for frequency regulation control. Using historical load data, time series forecasting algorithms, such as the ARIMA model, can be used to predict load demand for a period of time in the future. For example, based on the electricity consumption data of the past year, the system load for each hour in the next 24 hours is predicted. For example, the forecast results show that the system load will reach a peak of 5000MW at 9 am tomorrow. The frequency regulation capacity demand is calculated based on the transmission capacity limit threshold and the load forecast results, combined with the pre-set frequency regulation capacity estimation formula. For example, assuming that the system N-1 safety operation criteria require that the system can still maintain stable operation when any line fails. Combining load forecasting and line transmission capacity limits, it is calculated that the system needs to reserve 500MW of frequency regulation capacity. The multi-objective optimization model is established to optimize the power generation cost and spare capacity under various constraints, such as the upper and lower limits of the unit output, the ramp rate limit, etc. For example, the objective function of the model can be set to minimize the total power generation cost and the spare capacity cost, and the constraints include the output limit of each unit (for example, the minimum output is 100MW and the maximum output is 500MW) and the ramp rate limit (for example, the maximum ramp rate per minute is 10MW). By solving the model with a mixed integer programming solver, the optimal unit output and spare capacity configuration scheme can be obtained. For example, the output of a coal-fired unit is set to 400MW, the output of a hydropower unit is set to 200MW, and the system spare capacity is set to 500MW. According to the optimal configuration scheme, the active power amount that needs to be adjusted for each power station can be calculated. For example, the current output of a coal-fired power station is 350MW, and the optimization result is 400MW, then the power station needs to increase its output by 50MW. These adjustments will form active power adjustment instructions and be sent to the power station control system. After receiving the adjustment command, the power station control system adjusts the active power output of the power station in real time by adjusting the opening of the steam inlet valve of the steam turbine. For example, after receiving the command to increase the output by 50MW, the control system of the coal-fired power station will automatically increase the steam intake of the steam turbine, thereby increasing the speed of the generator and ultimately increasing the active power output. The grid frequency measurement device monitors the grid frequency in real time. If the frequency deviation exceeds the preset dead zone threshold (for example, ±0.03Hz), a new frequency modulation control process is triggered. For example, if the current grid frequency is 49.95Hz, which is lower than the lower limit of the dead zone threshold, the frequency modulation control process is triggered. The control system automatically adjusts the active power of the power station according to the direction of the frequency deviation and the preset frequency modulation rate until the frequency returns to the dead zone range.For example, if the grid frequency is lower than the lower limit of the dead zone threshold, the control system will automatically increase the active power output of some units until the grid frequency returns to between 49.97Hz and 50.03Hz. If the frequency still exceeds the threshold after frequency regulation, it means that there is a deviation between the previous load forecast and the frequency regulation capacity estimation. At this time, it is necessary to correct the load forecast model according to the actual frequency regulation results, re-forecast the load and calculate the frequency regulation capacity demand, and perform the next round of frequency regulation control until the frequency returns to stability. For example, if the grid frequency continues to decrease after the unit output is increased, it means that the actual load may be higher than the predicted value, and the load forecast needs to be re-performed and the frequency regulation capacity needs to be increased accordingly. Through the above process, real-time monitoring and automatic adjustment of the grid frequency can be achieved to ensure that the grid frequency is stable within the allowable range and to ensure the safe and stable operation of the power system.

[0032] Step S105: Perform safety verification analysis on the obtained optimized configuration scheme. Evaluate the impact of the scheme on the safe and stable operation of the power grid through power flow calculation and static safety analysis. If there are safety hazards such as overload and voltage exceeding the limit, return to the optimization model, adjust the corresponding constraints, and solve again.

[0033] Obtain the grid model parameters and operating condition data of the optimized configuration scheme, including network topology, line and transformer parameters, load power, etc., as input for power flow calculation and static safety analysis. Use the Newton-Raphson method for power flow calculation. According to the grid model parameters and operating condition data, calculate the power flow distribution of the grid under the optimized configuration scheme, and obtain the power flow and voltage level of each line and transformer. According to the static safety analysis criteria of the power system, determine whether the power flow calculation results have safety hazards such as line overload and voltage over-limit. If there are hidden dangers, the constraint adjustment mechanism is triggered. Through the pre-set optimization model constraint adjustment rules, the corresponding operating constraint parameters, such as line power limit and bus voltage limit, are automatically adjusted according to the type and severity of the safety hazard. The adjustment rules can be generated based on an expert knowledge base or a machine learning algorithm. Feedback the adjusted constraint parameters to the optimized configuration model, restart the solution process, and use optimization algorithms such as mixed integer programming to obtain a new optimized configuration scheme. Perform power flow calculation and static safety analysis on the new optimized configuration scheme again to evaluate whether the adjusted scheme meets the requirements for safe and stable operation of the power grid. If the new solution still has safety risks, return to the constraint adjustment step to continue iteration; if the new solution meets the requirements, it will be output as the final executable optimization configuration solution.

[0034] Specifically, obtaining the grid model parameters and operating condition data of the optimized configuration scheme is the basis for power flow calculation and static safety analysis. The grid model parameters include the network topology, such as which substations are connected by transmission lines; line and transformer parameters, such as line resistance, reactance, length, transformer ratio, capacity, etc.; load power, such as how much power load each substation is. The operating condition data includes real-time information such as generator output, load level, and line switch status. Together, these data form a complete picture of grid operation and are the basis for subsequent analysis. For example, a simple grid model with three substations can be obtained. Substations A and B are connected by a line, and substations B and C are also connected by a line. The line parameters are known, and the load power of each substation is also known. In addition, the output of the generator at substation A needs to be known. These data will be used as input for calculating power flow and performing safety analysis. The Newton-Raphson method is used for power flow calculation. The Newton-Raphson method is an iterative algorithm for solving nonlinear equations. In the power system, the goal of power flow calculation is to determine the voltage amplitude and phase angle of each node in the power grid, as well as the power flow of each line. Through the Newton-Raphson method, the power flow distribution of the power grid under the optimal configuration scheme can be calculated according to the power grid model parameters and operating condition data, and the power flow and voltage level of each line and transformer can be obtained. For example, in the example of the three substations mentioned above, the power flow of each line and the voltage amplitude and phase angle of each substation can be calculated by the Newton-Raphson method. After the power flow calculation results come out, static safety analysis is required. According to the static safety analysis criteria of the power system, it is judged whether there are safety hazards such as line overload and voltage over-limit in the power flow calculation results. Line overload means that the actual power flow of the line exceeds its maximum transmission capacity, which may cause the line to overheat or even damage. Voltage over-limit means that the voltage amplitude of the node exceeds the allowable range, which may affect the normal operation of the power equipment. For example, assuming that the maximum transmission capacity of line AB is 100MW, and the power flow calculation results show that the power flow of line AB is 120MW, then there is an overload risk in line AB. Assuming that the voltage allowable range of substation B is 0.95pu to 1.05pu, and the power flow calculation results show that the voltage of substation B is 0.9pu, the voltage of substation B is lower than the limit. If the static safety analysis finds that there is a safety hazard, the constraint adjustment mechanism needs to be triggered. Through the pre-set optimization model constraint adjustment rules, the corresponding operation constraint parameters, such as line power limit, bus voltage limit, etc., are automatically adjusted according to the type and severity of the safety hazard. The adjustment rules can be generated based on an expert knowledge base or a machine learning algorithm. For example, if line AB is overloaded, the power limit of line AB can be increased, or the output of substation A can be reduced. If the voltage of substation B is lower than the limit, the output of substation A can be increased, or the transformer ratio can be adjusted.Feedback the adjusted constraint parameters to the optimization configuration model and restart the solution process. New optimization configuration schemes can be obtained by using optimization algorithms such as mixed integer programming. For example, the output of each generator can be recalculated using a mixed integer programming algorithm to meet the adjusted constraints. The new optimization configuration scheme is subjected to power flow calculation and static safety analysis again to evaluate whether the adjusted scheme meets the requirements for safe and stable operation of the power grid. This is an iterative process. If the new scheme still has safety hazards, return to the constraint adjustment step to continue the iteration; if the new scheme meets the requirements, it is output as the final executable optimization configuration scheme. For example, if the new power flow calculation results show that the power flow of line AB is 90MW and the voltages of all nodes are within the allowable range, the scheme meets the requirements for safe and stable operation and can be output as the final scheme. This scheme can guide the actual operation of the power grid, such as adjusting the output of the generator to achieve optimal power configuration.

[0035] Step S106, according to the grid topology and line transformer parameters, the power flow and voltage level of the optimal configuration scheme are obtained by the power flow calculation method, and the static safety analysis is used to determine whether there are hidden dangers of line overload and voltage exceeding the limit. If so, the preset constraint adjustment rules are obtained, and the operation constraint parameters such as the line power limit and the bus voltage limit are automatically adjusted. The adjusted constraint parameters are fed back to the optimal configuration model for re-iteration and solution, and a new optimal configuration scheme is obtained. The power flow calculation and static safety assessment are performed again until the safe and stable operation requirements of the power grid are met.

[0036] According to the grid topology and line transformer parameters, a grid power flow calculation model is established in MATLAB / Simulink, and the Newton-Raphson method is used for power flow calculation to obtain the power flow and voltage level distribution under the initial optimal configuration scheme. By judging the limit of line power and bus voltage, it is judged whether the initial optimal configuration scheme has the hidden dangers of line overload and voltage over-limit. If there is no hidden danger, the current optimal configuration scheme is output as the final scheme, and the output content includes the voltage amplitude and phase angle of each bus, and the active power and reactive power of each line and transformer. If there is a hidden danger of line overload or voltage over-limit, the corresponding adjustment strategy is obtained from the preset constraint adjustment rule library. According to the adjustment strategy, the operating constraint parameters such as line power limit and bus voltage limit are automatically modified. The adjusted constraint parameters are used as new boundary conditions and substituted into the optimal configuration model. The sequential quadratic programming algorithm in the MATLAB optimization toolbox is used to solve the optimal configuration model to obtain a new optimal configuration scheme. The line power and bus voltage under the new scheme are used as the input conditions for power flow calculation, and the power flow calculation is performed again to obtain the power flow and voltage level distribution under the new optimal configuration scheme. According to the safety and stability limits of the power grid operation, the new optimization configuration scheme is statically checked for safety to determine whether the line power and bus voltage meet the limit requirements. If the requirements are met, the scheme is output as the final optimization configuration scheme; if the requirements are not met, the line power limit and bus voltage limit that do not meet the limit requirements under the new scheme are used as constraints, substituted into the constraint adjustment rule library for search and matching, and the corresponding adjustment strategy is obtained, and then the constraint parameter adjustment step is returned to continue iterative optimization and solution. Repeat the above iterative process of constraint parameter adjustment, optimization solution, power flow calculation and static safety check until the optimal configuration scheme that meets the requirements of safe and stable operation of the power grid is found. The bus voltage amplitude and phase angle, active power and reactive power of the line and transformer and other parameters of the scheme are output in a table form as the basis and guidance for the actual operation control of the power grid.

[0037] Specifically, power flow calculation is a basic tool in power system analysis, which is used to determine the steady-state operation state of the power grid under specific operating conditions. The goal of power flow calculation is to calculate the voltage amplitude and phase angle of each node in the power grid, as well as the power flow of each line. The following will combine MATLAB / Simulink and Newton-Raphson method for detailed description and examples. First, it is necessary to establish a power grid model in MATLAB / Simulink. Assume a simple three-node system, node 1 is a generator node, and nodes 2 and 3 are load nodes. Nodes 1 and 2, and nodes 2 and 3 are connected by transmission lines. Line parameters include resistance, reactance, and line length. In addition, it is necessary to know the load power of each node and the output of the generator. For example, the generator output of node 1 is 200MW, the load of node 2 is 100MW, and the load of node 3 is 80MW. The line parameters can be set as follows: the resistance of line 1-2 is 0.01pu and the reactance is 0.05pu; the resistance of line 2-3 is 0.02pu and the reactance is 0.06pu. These data form the basis for power flow calculation. The Newton-Raphson method is an iterative algorithm for solving nonlinear equations. In power flow calculation, the power balance equation needs to be solved. Through the Newton-Raphson method, it can be continuously iterated to gradually approach the final solution. The choice of initial value will affect the number of iterations and the accuracy of the final result. A reasonable initial value can speed up the convergence of the iteration. For example, the voltage amplitude of all nodes can be initialized to 1pu and the phase angle can be initialized to 0. Through power flow calculation, the voltage amplitude and phase angle of each node and the power flow of each line can be obtained. For example, after power flow calculation, the voltage amplitude of node 2 may be 0.98pu, the phase angle is -5 degrees, and the power flow of line 1-2 is 105MW. These results reflect the steady-state operation state of the power grid under given operating conditions. Next, static safety verification is required. The purpose of static safety verification is to evaluate whether the power grid is safe and stable under the current operating state. The main verification indicators include line power and bus voltage. The line power cannot exceed the maximum transmission capacity of the line, and the bus voltage must be within the allowable range. For example, assuming that the maximum transmission capacity of line 1-2 is 120MW, and the power flow calculation results show that the power flow of line 1-2 is 105MW, the operation status of line 1-2 is safe. Assuming that the allowable range of bus voltage is 0.95pu to 1.05pu, and the voltage amplitude of node 2 is 0.98pu, the voltage of node 2 is also within the allowable range. If the results of power flow calculation and static safety verification show that the operation status of the power grid is safe and stable, the current optimization configuration scheme can be output as the final scheme. The output content includes the voltage amplitude and phase angle of each bus, and the active power and reactive power of each line and transformer. These data can be used as a reference for the actual operation of the power grid.If there are safety hazards such as line overload or voltage over-limit, the constraints need to be adjusted and the optimization solution needs to be re-performed. For example, if line 1-2 is overloaded, the power transmission limit of line 1-2 can be increased, or the generator output of node 1 can be reduced. If the voltage of node 2 is too low, the generator output of node 1 can be increased, or the transformer ratio can be adjusted. The adjusted constraints will be used as new boundary conditions for solving the optimization configuration model. The optimization configuration model can be solved using the sequential quadratic programming algorithm in the MATLAB optimization toolbox. The sequential quadratic programming algorithm is an effective algorithm for solving nonlinear constrained optimization problems. By solving the optimization configuration model, a new optimization configuration scheme can be obtained. Substituting the new optimization configuration scheme into the power flow calculation model, a new power flow and voltage level distribution can be obtained. Then, a static safety check is performed again. If the new scheme meets the requirements for safe and stable operation, it is output as the final scheme. If it does not meet the requirements, the constraints need to be adjusted again, and the above iterative process is repeated until the optimal configuration scheme that meets the requirements is found. The bus voltage amplitude and phase angle of the final solution, active power and reactive power of the line and transformer and other parameters will be output in tabular form as the basis and guidance for the actual operation control of the power grid.

[0038] Step S107: After the optimized configuration scheme passes the safety check, a differentiated interconnection line power dispatching strategy is formulated in combination with the differences in section characteristics. For sections with severe load fluctuations, the power transmitted by the interconnection line is increased; for sections with relatively stable loads, the power transmitted is reduced to achieve a reasonable configuration of frequency modulation resources.

[0039] According to the configuration scheme, power flow calculation and static safety analysis are performed to determine whether the constraints such as voltage and line power flow are met. If so, the next step is executed, otherwise the configuration scheme is re-optimized. The historical load data of each section is obtained through the smart meter and power monitoring system, and the data preprocessing technology is used to clean the data and extract features to obtain the load fluctuation characteristic data. The sections are classified by the K-means clustering algorithm, and the sections are divided into two categories according to the load fluctuation characteristics: sections with severe load fluctuations and sections with stable loads. For sections with severe load fluctuations, the long short-term memory neural network (LSTM) algorithm is used. With the historical load data and the transmission power data of the tie line as input, a nonlinear mapping model of load fluctuation and transmission power is trained. The model is used to predict the future load fluctuation trend and determine the strategy of increasing the transmission power to this type of section under different load fluctuation conditions. For sections with stable loads, the CART decision tree algorithm is used. With the historical load data and the transmission power data of the tie line as input, the optimal splitting attribute is determined through information entropy and Gini index, and the decision rules for load stability and reduction of transmission power are generated to form a strategy for reducing the transmission power to this type of section. The strategy of increasing transmission power in sections with severe load fluctuations and the strategy of reducing transmission power in sections with stable loads are integrated to form a complete interconnection line power dispatching strategy, including transmission power adjustment plans for different section types and load fluctuations. The dispatching strategy is sent to the host computer of the interconnection line control system through the communication protocol, and the host computer controls the interconnection line power electronic equipment to change the interconnection line flow and implement the dispatching strategy. The interconnection line control system uses state estimation technology to dynamically evaluate the load fluctuation of the section based on the real-time load data of the section, and adjusts the transmission power of the interconnection line in real time according to the dispatching strategy to achieve the optimal configuration of frequency modulation resources among various sections. By real-time monitoring of indicators such as voltage, frequency, and active power of each section, the safety and stability of the power grid is evaluated. If safety and stability problems occur, protection control measures are triggered to remove faulty equipment to ensure safe and stable operation of the power grid.

[0040] Specifically, the rationality of the grid configuration scheme is directly related to the safe and stable operation of the grid. In order to ensure the effectiveness of the configuration scheme, it is necessary to perform power flow calculation and static safety analysis on it. Power flow calculation is to calculate the voltage of each node and the current of each branch of the grid according to the structure, parameters and operating conditions of the grid. Static safety analysis is to evaluate whether the grid meets the safety constraints such as voltage and line power flow under various operating conditions based on power flow calculation. For example, for a simple grid with three substations and two transmission lines, the power flow of each line and the voltage amplitude and phase angle of each substation can be obtained through power flow calculation. Then, through static safety analysis, it is determined whether the line power flow exceeds the transmission capacity of the line and whether the substation voltage is within the allowable range. If the calculation results show that the line is overloaded or the voltage exceeds the limit, the configuration scheme needs to be re-optimized, such as adjusting the transformer ratio or changing the output of the generator until all safety constraints are met. In order to achieve more refined interconnection line power scheduling, it is necessary to make full use of the historical load data obtained by smart meters and power monitoring systems. After preprocessing these data, including data cleaning and feature extraction, load fluctuation characteristic data can be obtained. Data cleaning is to remove noise and errors in the data. For example, abnormal load spikes can be removed. Feature extraction is to extract useful information from the raw data. For example, the variance and peak factor of the load data can be calculated to reflect the degree of load fluctuation. The K-means clustering algorithm can be used to classify sections according to the load fluctuation characteristics. The K-means algorithm is an unsupervised learning algorithm that can divide sections into two categories: sections with severe load fluctuations and sections with stable loads. For example, suppose there are two sections, one with large load fluctuations and the other with relatively stable loads. The K-means algorithm can divide these two sections into different categories. For sections with severe load fluctuations, the long short-term memory neural network (LSTM) algorithm can be used to establish a nonlinear mapping model between load fluctuations and transmission power. The LSTM algorithm is a recurrent neural network that is good at processing time series data. For example, the historical load data and the transmission power data of the tie line over a period of time can be used as the input of the LSTM network, and the network can be trained to learn the relationship between load fluctuations and transmission power, so as to predict future load fluctuation trends and formulate corresponding transmission power adjustment strategies. If it is predicted that the load is about to increase significantly, the transmission power of the tie line can be increased in advance to meet the power demand of the section. For sections with stable loads, the CART decision tree algorithm can be used to generate decision rules for load stability and reducing transmission power. The CART decision tree algorithm is a classification and regression tree algorithm that can build a decision tree model based on historical load data and tie line transmission power data. For example, if the load of a section remains stable for a long time, the decision tree model can determine that the section can reduce the transmission power of the tie line, thereby saving energy.By integrating the transmission power adjustment strategies for different section types, a complete tie-line power dispatching strategy can be formed. The strategy includes transmission power adjustment schemes for different section types and load fluctuations. For example, for sections with severe load fluctuations, the strategy may stipulate that the transmission power is dynamically adjusted according to the prediction results of the LSTM model; for sections with stable loads, the strategy may stipulate that the transmission power is reduced according to the judgment results of the CART decision tree model. The execution of the dispatching strategy depends on the tie-line control system. The dispatching strategy is sent to the host computer of the tie-line control system through the communication protocol, and the host computer controls the tie-line power electronic equipment to change the tie-line power flow. For example, the host computer can control the trigger angle of the converter to adjust the transmission power of the tie-line. The tie-line control system uses state estimation technology to dynamically evaluate the load fluctuation of the section and adjust the transmission power of the tie-line in real time according to the dispatching strategy. State estimation technology can estimate the state variables of the section, such as voltage and frequency, based on the real-time measurement data of the section. In order to ensure the safe and stable operation of the power grid, it is necessary to monitor the voltage, frequency, active power and other indicators of each section in real time. If safety and stability problems occur, such as voltage exceeding the limit or excessive frequency deviation, protection control measures need to be triggered to remove the faulty equipment. For example, if it is detected that the current of a line exceeds its rated value, the line needs to be disconnected immediately to prevent the line from overloading and causing a fault.

[0041] Step S108, establish a real-time monitoring and dispatching mechanism, deploy a wide-area measurement system, and collect the load level and transmission power data of each section in real time. Determine whether there is a large deviation. If the deviation exceeds the preset threshold, trigger the optimization model to recalculate and dynamically adjust the interconnection line power. The preset threshold is set based on historical data statistics and expert experience, and can be dynamically adjusted through an adaptive algorithm.

[0042] Obtain the real-time data of the load level and transmission power of each section, and transmit the data to the monitoring center server. Determine whether the load level and transmission power deviation of each section exceeds the preset threshold according to the preset threshold. If it exceeds the threshold, the optimization model is triggered. The optimization model is based on real-time data, historical data and expert experience, and obtains the optimal interconnection line power adjustment plan through support vector machine algorithm training. The adjustment plan calculated by the optimization model is sent to the dispatching system, and the interconnection line power is dynamically adjusted through remote control equipment. Continuously monitor the changes in the load level and transmission power of each section. If it exceeds the preset threshold in step 2 again, the optimization model is re-triggered to calculate the latest adjustment plan. The Kalman filter algorithm is used to dynamically adjust the preset threshold according to the actual operation situation to adapt to the changes in power grid load and optimization needs. Regularly evaluate and improve the optimization model, realize incremental learning through the gradient descent method, and use the newly collected data to update the model online to improve the model prediction accuracy and adjustment effect.

[0043] Specifically, in the power system, real-time monitoring of the load level and transmission power of each section is the key to ensuring the stable operation of the power grid. For example, suppose that the load level in the main commercial area of ​​a city rises sharply in the summer due to the increase in the use of air conditioners. The data received in real time by the monitoring center server shows that the transmission power in the area is rapidly approaching the upper limit of safe operation. At this time, the system automatically triggers the optimization model, which is based on the support vector machine algorithm and uses real-time data to compare and analyze historical data of the same period to predict the load change trend in the next few hours. The application of the support vector machine algorithm here is to find an optimal decision boundary to distinguish between sections that need to adjust the transmission power and sections that do not need to be adjusted. The algorithm generates a classification model by analyzing the load fluctuation pattern in the historical data and the current real-time data. This model can predict which sections may need to increase the transmission power due to excessive load. For example, if the forecast results show that the load in the commercial area continues to increase in the next two hours, the dispatching system will increase the transmission power to the area through remote control equipment according to the recommendations of the optimization model. In order to ensure the accuracy of the adjustment measures, the system will continuously monitor the actual effect of the adjustment. If the load level and transmission power in the commercial area approach the threshold again, the system will re-trigger the optimization model and adjust the transmission strategy according to the latest data. This dynamic adjustment mechanism can effectively respond to emergencies or unexpected load changes and ensure the stability and security of the power grid. In addition, in order to adapt to the long-term changes in the power grid load and the optimization needs, the system uses the Kalman filter algorithm to dynamically adjust the preset threshold. Kalman filtering is an effective prediction and correction method that can continuously update and optimize the prediction model based on actual measurement data. For example, in the past year, due to the construction of several large industrial parks, the power demand in a certain area has increased significantly. The Kalman filter algorithm can dynamically adjust the threshold according to these changes, making the system more sensitive and accurate in responding to actual needs. Finally, in order to continuously improve the performance of the optimization model, it is necessary to evaluate and improve the model regularly. By implementing incremental learning through the gradient descent method, the system can use the newly collected data to update the model online. This method not only improves the prediction accuracy of the model, but also enhances the adaptability and effectiveness of the adjustment strategy. For example, by analyzing the data of the past six months, it is found that the load fluctuations of certain sections in a specific time period are significantly different from the past. The system can adjust the model parameters to more accurately predict and respond to possible changes in the future.

[0044] Step S109, introduce scenario analysis and risk assessment methods, analyze the dynamic response characteristics of interconnection line power for typical scenarios of load mutation and power output fluctuation, and evaluate the adaptability of the solution. Based on this, the optimization model is improved to enhance the risk resistance of the solution. The scenario analysis and risk assessment results are used as input to guide the subsequent tracking evaluation and dynamic optimization process.

[0045] According to typical scenarios such as load mutation and power output fluctuation, historical and real-time data of interconnection line power are obtained. Digital filtering and wavelet transform are used to denoise and extract features of data, and key indicators reflecting the dynamic response characteristics of interconnection line power, such as overshoot, adjustment time and steady-state error, are obtained. Machine learning algorithms such as support vector regression or random forest are used, with scenario parameters (such as load mutation, power output fluctuation amplitude and frequency, etc.) as input and key indicators as output to establish a nonlinear mapping model between the dynamic response characteristics of interconnection line power and scenario parameters. Grid search and cross-validation are used to optimize model hyperparameters to improve the prediction accuracy and generalization ability of the model. For different load mutation and power output fluctuation scenarios, a series of scenario parameter combinations are randomly generated, and the established dynamic response model is used to predict the dynamic change trend of interconnection line power. Based on the prediction results, mean square error and dynamic performance indicators are used to evaluate the adaptability and risk resistance of the current optimization scheme. If the prediction results show that the current optimization scheme has insufficient adaptability or low risk resistance, then according to the prediction results and evaluation indicators, the objective function, constraint conditions, etc. of the optimization model are dynamically adjusted using intelligent optimization methods such as genetic algorithms or particle swarm optimization to generate an improved optimization scheme. The improved optimization scheme is converted into executable control instructions and sent to the tie line control and protection device through communication methods such as fieldbus or industrial Ethernet. At the same time, the real-time monitoring data of the tie line power is obtained through smart meters and data acquisition units, and preprocessed by methods such as data cleaning and anomaly detection. The real-time monitoring data is analyzed and pattern recognized using methods such as online sequence learning and concept drift detection to determine whether the actual operation effect of the optimization scheme meets the expected requirements. If the actual operation effect does not meet expectations, the real-time monitoring data and evaluation results are fed back to the optimization model to trigger the online learning and adaptive adjustment mechanism based on reinforcement learning to continuously optimize and improve the adaptability and risk resistance of the scheme. According to the results of tracking evaluation and dynamic optimization, a scenario analysis and risk assessment report is formed. The report provides the dynamic response curves of interconnection line power under different scenarios, the statistical distribution of key indicators, and the adaptability and risk resistance level of the optimization scheme, providing data support and theoretical basis for subsequent optimization decisions and scheme improvements, and realizing closed-loop management and continuous improvement of interconnection line power optimization.

[0046] Specifically, the core goal of the interconnection line power optimization scheme is to ensure the stable operation of the power system under various complex working conditions and maximize the energy utilization efficiency. The scheme forms a closed-loop management system through real-time monitoring, data analysis, predictive evaluation and adaptive adjustment. The following will elaborate on the various technical topics in the scheme and illustrate them with specific examples. First, data acquisition and preprocessing are the basis of the entire scheme. Deployment of a wide-area measurement system can collect load levels and transmission power data of each section in real time. For example, in a regional power grid, a large number of smart meters and sensors are installed to monitor the voltage, current, power and other parameters of each line in real time. Then, digital filtering and wavelet transform are used to denoise and extract features from the data. For example, if there is some high-frequency noise in the voltage data of a section, a low-pass filter can be used to filter it out to obtain a smoother voltage curve. At the same time, wavelet transform can also be used to extract the mutation components in the voltage signal to determine whether there is a sudden rise or fall in voltage. These key indicators, such as overshoot, adjustment time and steady-state error, will be used for subsequent model construction and evaluation. Secondly, building a dynamic response model is a key link in the scheme. Machine learning algorithms such as support vector regression or random forest can be used to establish a nonlinear mapping model between the dynamic response characteristics of the tie line power and the scenario parameters. For example, assuming that during the peak power consumption in a certain area in summer, a sudden increase in load will cause the tie line power to fluctuate. Historical data can be collected, including load mutation, power output fluctuation amplitude and frequency, tie line power overshoot, adjustment time and steady-state error. Use these data to train a support vector regression model, input scenario parameters, and predict key indicators. The hyperparameters of the model can be optimized through methods such as grid search and cross-validation to improve the prediction accuracy and generalization ability of the model. For example, the kernel function and penalty coefficient of the support vector machine model can be adjusted to find the best model parameter combination. Next, scenario prediction and solution evaluation are important components of the solution. For different load mutation and power output fluctuation scenarios, a series of scenario parameter combinations are randomly generated. For example, assuming that a tie line connects two cities, you can simulate the situation where a large-scale event in one of the cities causes a sudden increase in load, and set different load mutation amounts and durations. Use the established dynamic response model to predict the dynamic change trend of the tie line power. For example, predict whether the tie line power will exceed the safety threshold and whether the adjustment time meets the requirements. Based on the prediction results, the mean square error and dynamic performance indicators are used to evaluate the adaptability and risk resistance of the current optimization scheme. For example, if the prediction results show that the tie line power overshoot is too large, it means that the current optimization scheme has low risk resistance. If the evaluation results are not ideal, the scheme needs to be optimized and improved. Using intelligent optimization methods such as genetic algorithms or particle swarm optimization, the objective function, constraints, etc. of the optimization model are dynamically adjusted.For example, if the objective function of the current optimization scheme is to minimize the power fluctuation of the tie line, it can be adjusted to maximize the energy utilization efficiency under the premise of ensuring safety. The improved optimization scheme is converted into executable control instructions and sent to the tie line control and protection device. Finally, online learning and adaptive adjustment are the continuous improvement mechanisms of the scheme. Real-time monitoring data of the tie line power is obtained through smart meters and data acquisition units. The real-time monitoring data is analyzed and pattern recognized by using methods such as online sequence learning and concept drift detection to determine whether the actual operation effect of the optimization scheme meets the expected requirements. For example, the real-time changes of the tie line power can be monitored and compared with the predicted results. If the actual power fluctuation is much greater than the predicted value, it means that the model may have a deviation. At this time, the online learning and adaptive adjustment mechanism of reinforcement learning can be used to continuously optimize and improve the adaptability and risk resistance of the scheme based on real-time monitoring data and evaluation results. Finally, a scenario analysis and risk assessment report is formed to provide data support and theoretical basis for subsequent optimization decisions and scheme improvements.

[0047] Step S1010, track and evaluate the optimization configuration scheme and dynamically optimize it. Compare and analyze the deviation between the actual transmission of the interconnection line power and the optimization target, and identify the shortcomings of the scheme. According to the data characteristics and the complexity of the problem, select a support vector machine or a neural network algorithm to establish a mapping relationship between the deviation and the influencing factors. Support vector machines are suitable for small samples and high-dimensional data, and neural networks are suitable for large-scale, nonlinear problems. Use the established mapping relationship to provide a basis for dynamic optimization, continuously improve the rationality and effectiveness of the interconnection line power configuration, and ensure the safe and stable operation of the power grid.

[0048] The actual electric power transmission data and optimization target data of the tie line are obtained, and the data are preprocessed such as cleaning and normalization, and a suitable feature vector is constructed. The root mean square error between the actual electric power transmission data and the optimization target data is calculated as the deviation value to measure the degree of difference between the two. The complexity of the problem is judged according to the dimension, scale and other characteristics of the tie line electric power data. If the data dimension is high and the sample size is small, the support vector regression (SVR) algorithm is selected; if the data scale is large and there is a nonlinear relationship, the feedforward neural network (FFNN) algorithm is selected. The hyperparameters of the selected machine learning algorithm are optimized by grid search or random search, such as the kernel function type and penalty coefficient of SVR, or the number of hidden layers and neurons of FFNN. The deviation value is used as the target variable, and the various factors affecting the deviation are used as independent variables. The machine learning model is trained using the training set data, and the performance of the model is evaluated on the validation set, such as the root mean square error, mean absolute error and other indicators. The parameters of the configuration scheme are input into the trained machine learning model to obtain the predicted deviation value and identify the shortcomings of the scheme. For the identified deficiencies, heuristic optimization algorithms such as particle swarm optimization (PSO) are used to adjust the relevant parameters of the optimized configuration scheme, and then input into the machine learning model again for deviation prediction. The process of parameter adjustment and deviation prediction is repeated until the predicted deviation value is less than the set threshold or the maximum number of iterations is reached, thereby obtaining a dynamically optimized configuration scheme. The optimized configuration scheme is converted into control instructions and sent to the dispatching control equipment of the interconnection line through the energy management system (EMS) to realize the application of the optimized configuration scheme in the actual power grid operation. The actual operation effect of the optimized configuration scheme is continuously tracked and evaluated, and indicators such as line loss rate and voltage qualification rate before and after optimization are statistically analyzed to quantitatively evaluate the effectiveness of the optimization, and further improve the optimization model and algorithm based on the evaluation results.

[0049] Specifically, the actual power transmission data and optimization target data of the tie line are obtained. In order to optimize and control, it is necessary to collect the actual power transmission data of the tie line and the optimization target data preset according to the system requirements. For example, the actual transmission power may be the power value per minute monitored in real time, and the optimization target data may be the expected power value obtained based on the load forecast and economic dispatch plan. These data need to be cleaned, such as removing outliers and missing values, and normalized, such as scaling the data to between 0 and 1, to facilitate subsequent machine learning model processing. In addition, it is necessary to construct a suitable feature vector, such as taking time, meteorological data, load level, etc. as features to more comprehensively describe the state of the system. The root mean square error between the actual power transmission data and the optimization target data is calculated as the deviation value. The root mean square error is a common indicator to measure the degree of difference between the actual value and the target value. For example, if the actual transmission power is 100MW and the target power is 90MW, the deviation value is 10MW. By calculating the deviation value, the effectiveness of the current solution can be quantitatively evaluated. According to the characteristics of the dimension and scale of the tie line power data, the complexity of the problem can be judged. If the data dimension is high and the sample size is small, for example, there are only a few hundred samples, but each sample has dozens of features, then the support vector regression (SVR) algorithm is suitable because it performs well in processing high-dimensional small sample data. If the data scale is large and there are nonlinear relationships, for example, there are millions of samples, and there is a complex nonlinear relationship between power and other factors, then the feedforward neural network (FFNN) algorithm is suitable because it has strong nonlinear fitting capabilities. Use methods such as grid search or random search to optimize the hyperparameters of the selected machine learning algorithm. For the SVR algorithm, parameters such as the kernel function type and penalty coefficient need to be optimized. For example, radial basis kernel function or linear kernel function can be selected, and the best penalty coefficient can be found through grid search to improve the prediction accuracy of the model. For the FFNN algorithm, parameters such as the number of hidden layers and the number of neurons need to be optimized. For example, the best network structure can be determined through random search to balance the complexity and generalization ability of the model. Take the deviation value as the target variable and the various factors that affect the deviation as independent variables, and use the training set data to train the machine learning model. For example, time, meteorological data, load level, etc. can be used as independent variables, and the deviation value can be used as the target variable, and the SVR or FFNN model can be trained using historical data. Then, the performance of the model is evaluated on the validation set, such as calculating the root mean square error, mean absolute error and other indicators to select the model with the best performance. The parameters of the configuration scheme are input into the trained machine learning model to obtain the predicted deviation value and identify the shortcomings of the scheme. For example, the current tie line control parameters are input into the trained model to obtain the predicted deviation value. If the predicted deviation value is large, it means that the current scheme has shortcomings and needs to be optimized.For the identified deficiencies, heuristic optimization algorithms such as particle swarm optimization (PSO) are used to adjust the relevant parameters of the optimization configuration scheme. For example, if the model prediction deviation value is too large, the PSO algorithm can be used to adjust the control parameters of the tie line, such as changing the power setting value or adjusting the control strategy to reduce the deviation value. Repeat the process of parameter adjustment and deviation prediction until the prediction deviation value is less than the set threshold or the maximum number of iterations is reached. Through repeated iterations, the configuration scheme is continuously optimized until the prediction deviation value meets the requirements, thereby obtaining a dynamically optimized configuration scheme. The optimized configuration scheme is converted into control instructions. For example, the optimized power setting value and control strategy are converted into specific control instructions, which are sent to the dispatching control equipment of the tie line, such as circuit breakers or transformer tap changers, through the energy management system (EMS). The actual operation effect of the optimized configuration scheme is continuously tracked and evaluated. For example, the line loss rate and voltage qualification rate before and after the optimization are statistically analyzed to quantitatively evaluate the effectiveness of the optimization. If the evaluation result is not ideal, the optimization model and algorithm need to be further improved, such as adjusting the hyperparameters of the model or selecting a more appropriate algorithm. Continuous tracking and evaluation can form a closed-loop control, continuously improve the plan, and ensure that the communication line always operates in the best condition.

[0050] The description of the above embodiments is only used to help understand the technical solutions and core ideas of the present application. Ordinary technical personnel in this field should understand that they can still modify the technical solutions recorded in the aforementioned embodiments, or replace some of the technical features therein with equivalents. However, these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for coordinated control of power grid frequency regulation taking into account tie line sections, characterized in that: include: Obtaining load characteristic parameters, transmission capacity parameters and stability margin parameters of the section spanned by the tie line; A multi-objective optimization model is established according to the load characteristic parameters, transmission capacity parameters and stability margin parameters, wherein the objective function of the multi-objective optimization model is to minimize the mean square error of the interconnection line power, and the constraints include load prediction curve constraints, transmission capacity constraints, stability margin constraints, frequency regulation capacity constraints and overload capacity constraints; The multi-objective optimization model is solved by using an improved particle swarm optimization algorithm to obtain an optimal interconnection line power distribution curve that meets various constraints; Discretizing the optimal tie-line electric power distribution curve and converting it into an active power command value of a tie-line control system; The tie line control system adjusts the electric power of the tie line in real time according to the active power command value, thereby realizing optimized control of the tie line power flow.

2. The method according to claim 1, characterized in that The step of obtaining the load characteristic parameters of the section spanned by the tie line comprises: According to the historical load data of the tie line section, a load forecasting model is established using a long short-term memory neural network to obtain the load forecasting curve for a period of time in the future; According to the physical parameters and thermal stability limit of the tie line, a maximum transmission power threshold of the tie line is determined, and the threshold is used as a transmission capacity constraint condition of the multi-objective optimization model.

3. The method according to claim 2, characterized in that According to the historical load data of the tie line section, including: Use data cleaning and normalization preprocessing techniques to process the data and obtain a standardized historical load data set; Based on the preprocessed historical load data, a feature set is constructed and load-related features are selected as model input; The LSTM neural network algorithm is used to set the network structure parameters and training parameters, and the characteristic data is used for training and optimization to obtain the load prediction model. In the optimization power flow model with the goal of minimizing the power generation cost, the maximum transmission power constraint is used as the upper limit constraint of the interconnection line power flow. At the same time, the power balance constraint and the upper and lower limit constraints of the generator output are combined to construct the optimization model.

4. The method according to claim 1, characterized in that The step of using the improved particle swarm optimization algorithm to solve the multi-objective optimization model includes: Adaptive inertia weights and learning factors are introduced, and a non-dominated sorting strategy is adopted. By designing fitness functions and update strategies, the frequency regulation requirements and transmission capacity limitations of each section are dynamically balanced during the search process to obtain the optimal interconnection line power distribution curve that meets multiple constraints.

5. The method according to claim 1, characterized in that: According to the physical parameters of the line, the power flow calculation method is used to obtain the transmission capacity limit threshold of the line, obtain historical load data, and use time series prediction algorithms such as the ARIMA model to predict the load demand in the future; According to the transmission capacity limit threshold and load forecast results, the preset frequency regulation capacity estimation formula is used to calculate the required frequency regulation capacity demand; Establish a multi-objective optimization model with power generation cost and spare capacity as objective functions and upper and lower limits of unit output and ramp rate limit as constraints, and use a mixed integer programming solver to solve the model to obtain the optimal unit output and spare capacity configuration plan; The grid frequency measured in real time by the grid frequency measuring device is obtained. If the frequency deviation exceeds the dead zone threshold specified by the preset grid frequency regulation standard, a new round of frequency regulation control process is triggered.

6. The method according to claim 1, characterized in that After obtaining the optimal tie line power distribution curve, the method further includes: According to the electrical parameters of the tie line, a stability margin index of the optimal tie line electric power distribution curve is calculated by using a continuous power angle stability margin and a transient stability margin evaluation method; It is determined whether the stability margin index meets the stability requirement. If not, the stability margin index is used as a new constraint condition and returned to the multi-objective optimization model for re-solving.

7. The method according to claim 1, characterized in that The step of adjusting the electric power of the tie line in real time comprises: Obtaining real-time load data and transmission power data of each section of the tie line, and determining whether the deviation thereof exceeds a preset threshold; If it exceeds the preset threshold, the multi-objective optimization model is triggered to recalculate, generate a new active power command value, and adjust the electric power of the tie line until the deviation value is less than the preset threshold.

8. The method according to claim 1, characterized in that After obtaining the optimal tie line power distribution curve, the method further includes: Formulate differentiated interconnection line power dispatch strategies based on differences in cross-section characteristics; For sections with severe load fluctuations, increase the electric power transmitted by the tie line; For sections with relatively stable loads, reduce the transmission power and achieve reasonable allocation of frequency regulation resources.

9. The method according to claim 1, characterized in that The steps of establishing the multi-objective optimization model include: Taking the minimum mean square error of interconnection line power as the objective function, and taking load forecast curve, transmission capacity constraint, stability margin constraint, frequency regulation capability constraint and overload capability constraint as constraints, a multi-objective optimization model is constructed. The load prediction curve is obtained based on the historical data and prediction algorithm of the load of each section, the transmission capacity constraint is calculated based on the physical parameters of the interconnection line, the stability margin constraint is determined based on the electrical parameters of the interconnection line and the margin evaluation method, the frequency regulation capacity constraint is determined based on the frequency regulation capacity and load characteristics of the power grid, and the overload capacity constraint is determined based on the line conductor parameters and environmental conditions.

Images