Multi-protocol based dynamic control method for power distribution automation

Abstract

The application provides a power distribution automation dynamic control method based on multi-reduction, relates to the field of power distribution automation, and comprises the following steps: a random number injection current is generated by a random generation module according to standard data packets, a memristor differential equation is generated according to the random number injection current, chaotic sequence values are obtained by solving the memristor differential equation, and immune parameters are periodically obtained by the random generation module according to immune intervals and historical data sets; a state vector is substituted into an instruction reinforcement model to generate a control instruction, a current reward value of the control instruction is calculated, and if the current reward value is greater than a preset change threshold, the value of the immune interval is optimized. By using the above method, the generated control instruction is not only based on the current state, but also considers the historical data set, and the immune interval is continuously optimized through the current reward value mechanism, so that targeted random fluctuations for the same region and the same detection type are generated.

Description

Technical Field

[0001] This invention relates to the field of power distribution automation, specifically to a dynamic control method for power distribution automation based on multiple protocols. Background Technology

[0002] With the increasing penetration of distributed power sources and fluctuating loads in distribution networks, such as photovoltaic systems and electric vehicle charging stations, the problem of random fluctuations in key electrical quantities such as power, voltage, and frequency of the power grid is becoming increasingly significant. Traditional distribution network automation control strategies mostly rely on adjustment methods based on fixed thresholds or preset rules. This rigid control mode often exhibits a lagging response and lacks flexibility when facing fluctuations with random changes in amplitude and frequency. Furthermore, traditional voltage regulating equipment, due to mechanical inertia or fixed logic, struggles to achieve millisecond-level precise matching and adaptive suppression, easily leading to under-regulation or over-regulation, affecting power quality and equipment safety.

[0003] Existing control methods, whether classical PID control or deterministic model prediction-based control, all rely on the core logic of outputting control commands according to predetermined rules. When faced with scenarios involving frequent, sudden fluctuations not covered by training data or exhibiting strong nonlinear characteristics, these methods lack the necessary stochastic exploration capabilities in generating control commands. The system can only respond within known empirical rules and cannot explore better control strategies by introducing controlled random disturbances. Consequently, it is prone to getting trapped in local optima when dealing with unknown fluctuation patterns, resulting in slow convergence and even control strategy failure.

[0004] Therefore, there is an urgent need for a dynamic control method for power distribution automation to adapt to this randomness. Summary of the Invention

[0005] To overcome existing technical problems, this invention provides a dynamic control method for power distribution automation based on multiple protocols. The random disturbance method used in this method can suppress potential fluctuations in advance.

[0006] The present invention adopts the following technical solution.

[0007] The multi-protocol-based dynamic control method for power distribution automation is characterized by the following steps: The protocol type of the input data is determined by the protocol model, the transmission quality score of the input data is calculated by the network monitoring module, the analysis module standardizes the input data to obtain standardized data, and the standard data packets are obtained by combining the protocol type and transmission quality score. The standardized data is then integrated into the corresponding historical dataset. The random generation module generates random number injection current based on standard data packets, generates memristor differential equations based on random number injection currents, and obtains chaotic sequence values ​​by solving memristor differential equations. The random generation module periodically obtains immune parameters based on immune intervals and historical datasets. The analysis module constructs a state vector based on standard data packets, chaotic sequence values, and immune parameters. It then substitutes the state vector into the instruction reinforcement model to generate control instructions and calculates the current reward value of the control instructions. The control instructions are then integrated into the corresponding historical dataset. If the current reward value is greater than a preset change threshold, the value of the immune interval is optimized.

[0008] As a further improvement of the present invention, the specific steps for calculating the transmission quality score of the input data include: weighted summation of the transmission delay and packet loss rate of the input data to obtain the transmission quality score; Before determining the protocol type of the input data through the protocol model, the following steps are included: input data in the same region and of the same detection type are selected manually or automatically, and the selected input data are integrated into the corresponding fusion dataset to obtain multiple fusion datasets; The specific steps for standardizing input data to obtain standardized data include: data cleaning and format standardization of the input data to obtain standardized data; If the input data is in a certain fusion dataset, the fusion weight is calculated based on the transmission quality score of the input data in the fusion dataset. The fusion weight is then used to perform a weighted sum of multiple standardized data, and the calculated value is used to update the standardized data.

[0009] As a further improvement of the present invention, the analysis module includes a database for storing the historical dataset, and the database also includes a protocol mapping encoding table, wherein each protocol type has a unique basic code in the protocol mapping encoding table, and the basic code includes multiple numbers; The specific steps for generating a random number injection current based on standard data packets include: obtaining a first random number by multiplying the transmission quality score by a preset first weighting coefficient; The corresponding base code is selected based on the protocol type in the standard data packet. If there are multiple protocol types in the standard data packet, all corresponding base codes are selected and averaged. The average value is taken as the base code of the standard data packet. The last digit of the first random number is selected as the discard number. Select the first basic code in a cyclical order from left to right or right to left. The numbers are removed, the remaining numbers in the basic encoding are summed, and the sum is multiplied by a preset second weighting coefficient to obtain a second random number. The first random number and the second random number are added together to obtain the random number injection current.

[0010] As a further improvement of the present invention, the expression of the memristor differential equation specifically includes:

[0011] in, It is a memristor state variable. , and All are memristor constant factors. It is a random number injected current; The steps for obtaining chaotic sequence values ​​by solving the memristor differential equation specifically include: performing numerical integration according to the Euler method and a preset solution time step, and normalizing the numerical values ​​obtained by modulo operation to obtain chaotic sequence values.

[0012] As a further improvement of the present invention, the immune parameters include a first immune coefficient. Second immune coefficient And update timestamps; the random generation module has a preset set of value ranges corresponding to each protocol type. The specific steps by which the random generation module periodically obtains immune parameters based on the immunization interval and historical dataset include: If this is the first time the immune parameters have been obtained, the first immune coefficient is randomly selected from the corresponding value range set. Second immune coefficient The value is then updated, and the update timestamp is updated accordingly; If this is not the first time the immune parameters have been acquired, and the time difference between the current time and the update timestamp is greater than the immune interval, then multiple first immune coefficients are randomly generated according to the corresponding value range set. and multiple factors related to the first immune coefficient One-to-one corresponding second immune coefficient The first immune coefficient of each pair Second immune coefficient Substituting the values ​​into the objective function yields the long-term error value, which is then used to calculate the first immunity coefficient of the pair. Second immune coefficient Long-term performance on historical datasets; The value with the smallest long-term error is selected as the center value, and the first immune coefficient corresponding to this center value is selected. Second immune coefficient According to the first immune coefficient Second immune coefficient The first immunity coefficient is obtained by performing iterative minimization calculation. Second immune coefficient The value is then updated, and the update timestamp is updated.

[0013] As a further improvement of the present invention, the specific steps of the iterative minimization calculation include: The first immune coefficient corresponding to the center value Second immune coefficient Multiple pairs of first immune coefficients are generated iteratively around the center. Second immune coefficient Each pair of first immune coefficients Second immune coefficient Substitute the values ​​into the objective function to calculate the current long-term error value. If there is a current long-term error value that is less than the center value, then record the current long-term error value with the smallest value as the center value, and repeat this step until the difference between the current long-term error value and the center value is less than the preset convergence threshold.

[0014] As a further improvement of the present invention, while determining the protocol type of the input data through the protocol separation model, the detection type of the input data is also determined through the protocol separation model. The instruction enhancement model has a pre-set instruction control specification library, which includes multiple standard control target values ​​that correspond one-to-one with the detection type. The specific steps for substituting the state vector into the instruction enhancement model to generate control instructions include: calculating the difference between the standard control target value and the value in the standardized data to obtain the basic action value; and then, based on the basic action value and the first immunity coefficient... The final action value is calculated based on the second immune coefficient. The system calculates the base noise value based on the preset noise scale, calculates the final noise value based on the base noise value and the chaotic sequence value, adds the final action value and the final noise value to obtain the final control value, and generates control commands.

[0015] As a further improvement of the present invention, the specific steps for calculating the current reward value of the control command include: calculating the action square based on the basic action value, calculating the fluctuation square based on the final noise value, and detecting whether the action square is greater than the safety change threshold corresponding to the detection type. If so, then generate the violation penalty coefficient, and sum the action square, fluctuation square, and violation penalty coefficient according to their weights to obtain the current reward value; If not, the current reward value is obtained by summing the squared action and squared fluctuation according to their weights.

[0016] As a further improvement of the present invention, the specific steps for optimizing the value of the immune interval include: generating a reduction coefficient that is inversely proportional to the current reward value based on the current reward value, wherein the reduction coefficient is less than 1, and multiplying the value of the immune interval by the reduction coefficient.

[0017] The beneficial effects of this invention are as follows: Random number injection current is generated using standard data packets. Chaotic sequence values ​​are indirectly obtained based on the random number injection current. Immune parameters are periodically acquired based on historical datasets. Then, a state vector is constructed using the standard data packets, chaotic sequence values, and immune parameters. Historical datasets, transmission quality scores, and chaotic perturbations are deeply integrated. The generated control commands are not only based on the current state but also comprehensively consider historical datasets. Furthermore, the immune interval is continuously optimized through a current reward value mechanism, thereby generating targeted random fluctuations for the same region and the same detection type. If the current reward value exceeds a change threshold, the system automatically adjusts the immune interval value to avoid over-adjustment that would cause the control commands within that immune interval to be inconsistent with the actual fluctuation situation. Moreover, random perturbations can be explored to help the system try new control strategies, thereby indirectly suppressing potential fluctuations in advance and preventing malfunctions. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the present invention. Detailed Implementation

[0020] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the scope of this patent. To better illustrate this embodiment, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual dimensions of the product.

[0021] It will be understood by those skilled in the art that certain well-known structures and their descriptions may be omitted in the accompanying drawings. The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0022] Reference Figure 1 This invention proposes a dynamic control method for power distribution automation based on multiple protocols, characterized by the following steps: The protocol type of the input data is determined by the protocol model, the transmission quality score of the input data is calculated by the network monitoring module, the analysis module standardizes the input data to obtain standardized data, and the standard data packets are obtained by combining the protocol type and transmission quality score. The standardized data is then integrated into the corresponding historical dataset. As a further improvement of the present invention, the specific steps for calculating the transmission quality score of the input data include: weighted summation of the transmission delay and packet loss rate of the input data to obtain the transmission quality score; Before determining the protocol type of the input data through the protocol model, the following steps are included: input data in the same region and of the same detection type are selected manually or automatically, and the selected input data are integrated into the corresponding fusion dataset to obtain multiple fusion datasets; The specific steps for standardizing input data to obtain standardized data include: data cleaning and format standardization of the input data to obtain standardized data; If the input data is in a certain fusion dataset, the fusion weight is calculated based on the transmission quality score of the input data in the fusion dataset. The fusion weight is then used to perform a weighted sum of multiple standardized data, and the calculated value is used to update the standardized data.

[0023] Because sensors exhibit inherent fluctuations, multiple sensors, even different models, may be installed in the same area for the same detection type. In such cases, generating control commands individually for each sensor can lead to conflicting commands. Conversely, fusing these commands increases computational complexity and amplifies the influence of randomness, magnifying fluctuations several times over. Therefore, integrating this data into a fused dataset treats the sensors as a whole, avoiding conflicting control commands and simplifying subsequent calculations. Furthermore, the fused dataset assists in generating the base code for the second random number, improving both security and randomness.

[0024] Because different sensors have different detection accuracies and transmission performance, the reliability of the detected values ​​is greatly reduced when the sensor transmission delay is large or even packet loss occurs. Therefore, the fusion weight is calculated by using the transmission quality score, and the weighted sum of all standardized data in the fusion dataset is performed based on the fusion weight. This can increase the numerical weight of sensors with better performance and decrease the numerical weight of sensors with poorer performance.

[0025] At this point, it's important to note that since different input data have different timestamps, aligning the timestamps for each input data point would be computationally and physically demanding. Therefore, as a specific embodiment of the present invention, the fused dataset can also be formed at different time intervals, such as one fused dataset per minute. In this case, the workload of aligning the timestamps of different input data can be reduced or even eliminated.

[0026] The expression for calculating the transmission quality score is as follows:

[0027] in, It is the transmission quality score of the i-th input data. and It is a quality weighting factor. This is the packet loss rate of the input data. It is the transmission delay of the input data. and All of these are minimum constants.

[0028] The expression for weighted summation of multiple standardized data points based on the fusion weights is as follows:

[0029] in, It is the fusion weight of the i-th input data. It is the smallest transmission quality score in this fused dataset. It is a fusion weight amplification factor. It is the total number of input data in this fused dataset. It is the value of the standardized data of the i-th input data.

[0030] also, The normalized weighted method can also be used for calculation.

[0031] Quality weighting factor The value must be greater than the quality weight factor. The reason for this high value is that packet loss rate, compared to transmission delay, is a more "serious" transmission error, and its probability of occurrence is very low. Normally, the packet loss rate is very low; in this case, without a quality weighting factor... and quality weighting factor Adjusting the ratio between the two leads to the packet loss rate metric almost completely determining the transmission quality score. This results in sensors with high transmission latency but near-zero packet loss rate having extremely high weights, distorting the standardized data detection values. More specifically, the packet loss rate refers to the packet loss rate over 24 hours, which can also be adjusted based on the transmission frequency of different sensors. The transmission quality score is converted into fusion weights using an exponential term because the packet loss rate and transmission latency of different sensors often don't differ significantly, thus requiring an exponential term for scaling.

[0032] In order to make the quality weighting factor and quality weighting factor The settings have high constraint specifications, therefore the constraints + =1. As a specific embodiment of the present invention, the quality weighting factor... It is 0.6, the quality weighting factor. It is 0.4, the fusion weight amplification factor. The value is 1.5. Suppose there are two sensors, A and B, used for voltage monitoring in the same area. Sensor A has a packet loss rate of 2%, a transmission delay of 0.05 seconds, and a detected value of 230V. Sensor B has a packet loss rate of 3%, a transmission delay of 0.04 seconds, and a detected value of 235V. The calculated transmission quality score for sensor A is 30 + 8 = 38, and for sensor B it is 19.999 + 10 = 29.999. The calculated fusion weight for sensor A is 1.534, and for sensor B it is 1.218. The weighted sum is then calculated to obtain a value of (230...). 1.534+235 1.218) / (1.534+1.218)=639.05 / 2.752=232.21V.

[0033] The random generation module generates random number injection current based on standard data packets, generates memristor differential equations based on random number injection currents, and obtains chaotic sequence values ​​by solving memristor differential equations. The random generation module periodically obtains immune parameters based on immune intervals and historical datasets.

[0034] As a further improvement of the present invention, the analysis module includes a database for storing the historical dataset, and the database also includes a protocol mapping encoding table, wherein each protocol type has a unique basic code in the protocol mapping encoding table, and the basic code includes multiple numbers; The specific steps for generating a random number injection current based on standard data packets include: obtaining a first random number by multiplying the transmission quality score by a preset first weighting coefficient; The corresponding base code is selected based on the protocol type in the standard data packet. If there are multiple protocol types in the standard data packet, all corresponding base codes are selected and averaged. The average value is taken as the base code of the standard data packet. The last digit of the first random number is selected as the discard number. Select the first basic code in a cyclical order from left to right or right to left. The numbers are removed, the remaining numbers in the basic encoding are summed, and the sum is multiplied by a preset second weighting coefficient to obtain a second random number. The first random number and the second random number are added together to obtain the random number injection current.

[0035] The transmission quality score of the input data is generally different each time. Generating a first random number based on this score allows for the formation of an unpredictable random number in a pseudo-random manner. Combining this with a second random number generates an almost completely random number. To prevent others from obtaining packet loss rate and transmission delay data through other means to calculate the first random number, this invention also includes a basic encoding scrambling mechanism. This mechanism shuffles several numbers in the basic encoding at regular intervals, causing them to be out of order. Even with the first random number known, it's impossible to determine which number in the basic encoding needs to be deleted, thus preventing the calculation of the second random number and resulting in an unpredictable and highly secure random number injection current.

[0036] The formulas for calculating the first and second random numbers are as follows:

[0037] in, It is the first random number. It is the second random number. It is the transmission quality score. It is the first weighting coefficient. It is the summation of the remaining numbers in the basic encoding. It is the second weighting coefficient.

[0038] Specifically, the first weighting coefficient Second weighting coefficient Both can be 1.

[0039] As a further improvement of the present invention, the expression of the memristor differential equation specifically includes:

[0040] in, It is a memristor state variable. , and All are memristor constant factors. It is a random number injected current; The steps for obtaining chaotic sequence values ​​by solving the memristor differential equation specifically include: performing numerical integration according to the Euler method and a preset solution time step, and normalizing the numerical values ​​obtained by modulo operation to obtain chaotic sequence values.

[0041] Specifically, The value is 1. The value is 1. The value is 0.001, random number injection current. The value is 1125, the solution time step is 0.01 seconds, and the first iteration... The value is 0.1. In this case, within the same 5 solution time steps, the obtained value is 0.16238309. Of course, in actual operation, there are far more than 5 iterations; this is only to illustrate the general calculation process and principle. Naturally, the specific parameters should be determined according to the actual situation. When you want fewer calculation steps and faster speed, you can increase the solution time step, but this will correspondingly reduce the accuracy. Since solving differential equations and the Euler method are common methods for those skilled in the art, they will not be elaborated upon further.

[0042] As a further improvement of the present invention, the immune parameters include a first immune coefficient. Second immune coefficient And update timestamps; the random generation module has a preset set of value ranges corresponding to each protocol type. The specific steps by which the random generation module periodically obtains immune parameters based on the immunization interval and historical dataset include: If this is the first time the immune parameters have been obtained, the first immune coefficient is randomly selected from the corresponding value range set. Second immune coefficient The value is then updated, and the update timestamp is updated accordingly; If this is not the first time the immune parameters have been acquired, and the time difference between the current time and the update timestamp is greater than the immune interval, then multiple first immune coefficients are randomly generated according to the corresponding value range set. and multiple factors related to the first immune coefficient One-to-one corresponding second immune coefficient The first immune coefficient of each pair Second immune coefficient Substituting the values ​​into the objective function yields the long-term error value, which is then used to calculate the first immunity coefficient of the pair. Second immune coefficient Long-term performance on historical datasets; The value with the smallest long-term error is selected as the center value, and the first immune coefficient corresponding to this center value is selected. Second immune coefficient According to the first immune coefficient Second immune coefficient The first immunity coefficient is obtained by performing iterative minimization calculation. Second immune coefficient The value is then updated, and the update timestamp is updated.

[0043] At this point, it is important to note the first immune coefficient. This corresponds to the gain, the second immune coefficient. This corresponds to thickness. Therefore, for different protocol types, both the gain range and thickness range need to be limited to avoid excessive divergence that could lead to control command errors. For example, some high-latency protocols tend to generate more conservative parameters, namely the first immunity coefficient. Smaller, second immune coefficient A larger value is used to enhance system stability and avoid control oscillations caused by communication and control delays. For some highly reliable protocols, more aggressive parameters can be generated, such as the first immunity coefficient. Larger, second immune coefficient A smaller value is needed to improve the overall system response speed. It is important to note at this point that the first immune coefficient... Second immune coefficient There is no direct or inverse proportional relationship between them; they are two independent data sets, and different first immunity coefficients may be generated based on different historical datasets. Second immune coefficient The value.

[0044] The objective function is expressed as follows:

[0045] in, It is the total number of recent control commands in the historical dataset. This is the first immune coefficient Second immune coefficient The value is the actual error value generated in the l-th control command.

[0046] More specifically, the total number of control commands can be set to the total number within one or two days. If a longer-term historical data performance is desired, it can be set to select all data from the past historical dataset and randomly select only a few data points from each day of the history based on the timestamp for calculation. However, due to differences in network conditions and specific user usage, the objective function is best calculated based on the most recent period, and the historical dataset is used to distinguish between two primary immunity coefficients: positive and negative deviations. Second immune coefficient The calculation.

[0047] The calculation of this actual error value involves using the first immunity coefficient. Second immune coefficient The numerical values ​​are substituted into the calculation of the final control values ​​mentioned later. Therefore, each time a control command is generated, it is integrated into the corresponding historical dataset to facilitate future updates to the first immunity coefficient. Second immune coefficient Substitute values ​​into the calculations to assess the first immunization coefficient of the pair. Second immune coefficient The long-term performance of [the product / service]. Since the formula is clear and simple, requiring only substitution to calculate the difference, squaring it, and finally summing, this invention will not elaborate on specific examples.

[0048] As a further improvement of the present invention, the specific steps of the iterative minimization calculation include: The first immune coefficient corresponding to the center value Second immune coefficient Multiple pairs of first immune coefficients are generated iteratively around the center. Second immune coefficient Each pair of first immune coefficients Second immune coefficient Substitute the values ​​into the objective function to calculate the current long-term error value. If there is a current long-term error value that is less than the center value, then record the current long-term error value with the smallest value as the center value, and repeat this step until the difference between the current long-term error value and the center value is less than the preset convergence threshold.

[0049] Specifically, the convergence threshold is set to 0.01, but it can also be set according to the desired system response speed and computational accuracy to avoid computational overload due to excessive iterations. This allows us to obtain a pair of first immunity coefficients. Second immune coefficient The values ​​that performed well in the historical dataset were used to generate this control command.

[0050] The analysis module constructs a state vector based on standard data packets, chaotic sequence values, and immune parameters. It then substitutes the state vector into the instruction reinforcement model to generate control instructions and calculates the current reward value of the control instructions. The control instructions are then integrated into the corresponding historical dataset. If the current reward value is greater than a preset change threshold, the value of the immune interval is optimized.

[0051] Random number injection currents are generated using standard data packets. Chaotic sequence values ​​are indirectly obtained from these injections, and immune parameters are periodically acquired based on historical datasets. A state vector is then constructed using the standard data packets, chaotic sequence values, and immune parameters. Historical datasets, transmission quality scores, and chaotic perturbations are deeply integrated, resulting in control commands that are not only based on the current state but also consider historical data. Furthermore, the immune interval is continuously optimized using a current reward value mechanism, thereby generating targeted random fluctuations for the same region and detection type. If the current reward value exceeds a change threshold, the system automatically adjusts the immune interval to avoid over-adjustment that could cause control commands within that interval to be inconsistent with actual fluctuations. The random perturbations can also be explored to help the system try new control strategies, thereby indirectly suppressing potential fluctuations in advance and preventing malfunctions.

[0052] As a further improvement of the present invention, while determining the protocol type of the input data through the protocol separation model, the detection type of the input data is also determined through the protocol separation model. Determining the protocol type is a standard technical method for those skilled in the art, such as through the protocol fields in the IP packet header or the file header, and therefore will not be elaborated upon further. As for determining the detection type, a standard feature semantic library can be manually set up, and the type can be determined directly by comparing the identified data content.

[0053] The instruction enhancement model has a pre-set instruction control specification library, which includes multiple standard control target values ​​that correspond one-to-one with the detection type. The specific steps for substituting the state vector into the instruction enhancement model to generate control instructions include: calculating the difference between the standard control target value and the value in the standardized data to obtain the basic action value; and then, based on the basic action value and the first immunity coefficient... The final action value is calculated based on the second immune coefficient. The system calculates the base noise value based on the preset noise scale, calculates the final noise value based on the base noise value and the chaotic sequence value, adds the final action value and the final noise value to obtain the final control value, and generates control commands.

[0054] The final expression for the control value is as follows:

[0055]

[0056] in, It is the final control value. It is the final action value. This is the final noise level. These are the basic action values. It is the first immune coefficient. It is the standard control target value. These are the test values ​​in the standardized data. This is the final noise level. It is a chaotic sequence value. It is the second immune coefficient. It is the preset noise level.

[0057] The final control value is divided into an action component and a noise component. The action component introduces a first immunity coefficient. Gain amplifies or reduces the amplitude of the action, and a second immunity coefficient is introduced into the noise component. The magnitude of random variation in thickness increase or decrease, and the first immune coefficient Second immune coefficient This was determined through validation using historical datasets, enabling optimal control. For example, when the standardized data values ​​for a certain detection type fluctuate significantly, it can negatively impact its secondary immunity coefficient. A larger value, meaning a larger range of random fluctuations, indicates better performance over long-term testing. In this case, the secondary immunity coefficient will also be increased when generating subsequent control commands. The numerical value of this increases the magnitude of random fluctuations. For example, standardized data for a certain type of test may surge at certain times, then rapidly decline and stabilize. In this case, the first immune coefficient... This will be relatively small, and less than 1, thus reducing the overall final control value.

[0058] For example, if the target voltage is 1.0 pu and the current measured voltage is 0.98 u (a positive deviation indicates that the current voltage is lower than the target and needs to be increased, such as by injecting reactive power through an inverter), the first immunity coefficient... It is 1.2, the second immune coefficient. The value is 0.2, the preset noise level is 0.1 pu, the chaotic sequence value is 0.3, the calculated basic action value is 0.02 pu, the final action value is 0.024 pu, the basic noise value is 0.02 pu, the final noise value is 0.006 pu, the calculated final control value is 0.03 pu, and the adjustment value in the control command is equal to... The specific value of the current measured voltage, such as 322V. 1.03 = 9.66V.

[0059] As a further improvement of the present invention, the specific steps for calculating the current reward value of the control command include: calculating the action square based on the basic action value, calculating the fluctuation square based on the final noise value, and detecting whether the action square is greater than the safety change threshold corresponding to the detection type. If so, then generate the violation penalty coefficient, and sum the action square, fluctuation square, and violation penalty coefficient according to their weights to obtain the current reward value; If not, the current reward value is obtained by summing the squared action and squared fluctuation according to their weights.

[0060] The square of the action reflects the intensity of the control action, while the square of the fluctuation reflects the uncertainty of the system. The greater the control action and the greater the uncertainty, the higher the risk, especially when the square of the action exceeds a predetermined safety change threshold. This significantly increases the danger, and consequently, when the current reward value exceeds the safety change threshold, the corresponding immune interval should be adjusted. Specifically, if the safety adjustment range for the current detection type is set to be no more than 3%, this safety change threshold can be set to be slightly less than the square of the standard control target value multiplied by 3%. Specifically, in the formula for calculating the current reward value, the weight in the weighted summation can be directly equal to 1. The standard control target value is dynamically adjusted according to the power grid status; the determination of its specific value is common in existing technologies, and therefore will not be elaborated upon further in this invention.

[0061] As a further improvement of the present invention, the specific steps for optimizing the value of the immune interval include: generating a reduction coefficient that is inversely proportional to the current reward value, wherein the reduction coefficient is less than 1, and multiplying the value of the immune interval by the reduction coefficient. The specific expression of the reduction coefficient formula is as follows:

[0062] in, It is a reduction factor. It refers to the specific reduction factor or magnification factor. This is the current reward value. It is the safety change threshold.

[0063] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A dynamic control method for power distribution automation based on multiple protocols, characterized in that, Includes the following steps: The protocol type of the input data is determined by the protocol model, the transmission quality score of the input data is calculated by the network monitoring module, the analysis module standardizes the input data to obtain standardized data, and the standard data packets are obtained by combining the protocol type and transmission quality score. The standardized data is then integrated into the corresponding historical dataset. The random generation module generates random number injection current based on standard data packets, generates memristor differential equations based on random number injection currents, and obtains chaotic sequence values ​​by solving memristor differential equations. The random generation module periodically obtains immune parameters based on immune intervals and historical datasets. The analysis module constructs a state vector based on standard data packets, chaotic sequence values, and immune parameters. It then substitutes the state vector into the instruction reinforcement model to generate control instructions and calculates the current reward value of the control instructions. The control instructions are then integrated into the corresponding historical dataset. If the current reward value is greater than a preset change threshold, the value of the immune interval is optimized.

2. The dynamic control method for distribution automation based on multiple protocols according to claim 1, characterized in that, The specific steps for calculating the transmission quality score of the input data include: weighting and summing the transmission delay and packet loss rate of the input data to obtain the transmission quality score; Before determining the protocol type of the input data through the protocol model, the following steps are included: input data in the same region and of the same detection type are selected manually or automatically, and the selected input data are integrated into the corresponding fusion dataset to obtain multiple fusion datasets; The specific steps for standardizing input data to obtain standardized data include: data cleaning and format standardization of the input data to obtain standardized data; If the input data is in a certain fusion dataset, the fusion weight is calculated based on the transmission quality score of the input data in the fusion dataset. The fusion weight is then used to perform a weighted sum of multiple standardized data, and the calculated value is used to update the standardized data.

3. The dynamic control method for power distribution automation based on multiple protocols according to claim 2, characterized in that, The analysis module includes a database for storing the historical dataset, and the database also includes a protocol mapping encoding table. Each protocol type has a unique basic code in the protocol mapping encoding table, and the basic code includes multiple numbers. The specific steps for generating a random number injection current based on standard data packets include: obtaining a first random number by multiplying the transmission quality score by a preset first weighting coefficient; The corresponding base code is selected based on the protocol type in the standard data packet. If there are multiple protocol types in the standard data packet, all corresponding base codes are selected and averaged. The average value is taken as the base code of the standard data packet. The last digit of the first random number is selected as the discard number. Select the first basic code in a cyclical order from left to right or right to left. The numbers are removed, the remaining numbers in the basic encoding are summed, and the sum is multiplied by a preset second weighting coefficient to obtain a second random number. The first random number and the second random number are added together to obtain the random number injection current.

4. The dynamic control method for distribution automation based on multiple protocols according to claim 1, characterized in that, The specific expression of the memristor differential equation includes: ， in, It is a memristor state variable. , and All are memristor constant factors. It is a random number injected current; The steps for obtaining chaotic sequence values ​​by solving the memristor differential equation specifically include: performing numerical integration according to the Euler method and a preset solution time step, and normalizing the numerical values ​​obtained by modulo operation to obtain chaotic sequence values.

5. The dynamic control method for distribution automation based on multiple protocols according to claim 1, characterized in that, Immune parameters include the first immune coefficient. Second immune coefficient And update timestamps; the random generation module has a preset set of value ranges corresponding to each protocol type. The specific steps by which the random generation module periodically obtains immune parameters based on the immunization interval and historical dataset include: If this is the first time obtaining immune parameters, the first immune coefficient is randomly selected from the corresponding value range set. Second immune coefficient The value is then updated, and the update timestamp is updated accordingly; If this is not the first time the immune parameters have been acquired, and the time difference between the current time and the update timestamp is greater than the immune interval, then multiple first immune coefficients are randomly generated based on the corresponding value range set. and multiple factors related to the first immune coefficient One-to-one corresponding second immune coefficient The first immune coefficient of each pair Second immune coefficient Substitute the values ​​into the objective function to obtain the long-term error value, and then calculate the first immunity coefficient. Second immune coefficient Long-term performance on historical datasets; The value with the smallest long-term error is selected as the center value, and the first immune coefficient corresponding to this center value is selected. Second immune coefficient According to the first immune coefficient Second immune coefficient The first immunity coefficient is obtained by performing iterative minimization calculation. Second immune coefficient The value is then updated, and the update timestamp is updated.

6. The dynamic control method for distribution automation based on multiple protocols according to claim 5, characterized in that, The specific steps of the iterative minimization calculation include: The first immune coefficient corresponding to the center value Second immune coefficient Multiple pairs of first immune coefficients are generated iteratively around the center. Second immune coefficient Each pair of first immune coefficients Second immune coefficient Substitute the values ​​into the objective function to calculate the current long-term error value. If there is a current long-term error value that is less than the center value, then record the current long-term error value with the smallest value as the center value, and repeat this step until the difference between the current long-term error value and the center value is less than the preset convergence threshold.

7. The dynamic control method for distribution automation based on multiple protocols according to claim 6, characterized in that, While determining the protocol type of the input data through the protocol-specific model, the detection type of the input data is also determined through the protocol-specific model. The instruction enhancement model has a pre-set instruction control specification library, which includes multiple standard control target values ​​that correspond one-to-one with the detection type. The specific steps for substituting the state vector into the instruction enhancement model to generate control instructions include: calculating the difference between the standard control target value and the value in the standardized data to obtain the basic action value; and then, based on the basic action value and the first immunity coefficient... The final action value is calculated based on the second immune coefficient. The system calculates the base noise value based on the preset noise scale, calculates the final noise value based on the base noise value and the chaotic sequence value, adds the final action value and the final noise value to obtain the final control value, and generates control commands.

8. The dynamic control method for distribution automation based on multiple protocols according to claim 7, characterized in that, The specific steps for calculating the current reward value of the control command include: calculating the action square based on the basic action value, calculating the fluctuation square based on the final noise value, and detecting whether the action square is greater than the safety change threshold corresponding to the detection type. If so, then generate the violation penalty coefficient, and sum the action square, fluctuation square, and violation penalty coefficient according to their weights to obtain the current reward value; If not, the current reward value is obtained by summing the squared action and squared fluctuation according to their weights.

9. The dynamic control method for distribution automation based on multiple protocols according to claim 1, characterized in that, The specific steps for optimizing the value of the immune interval include: generating a reduction coefficient that is inversely proportional to the current reward value, wherein the reduction coefficient is less than 1, and multiplying the value of the immune interval by the reduction coefficient.

Images