Multi-energy collaborative zero-carbon park smart energy system based on AI intelligent agent autonomous regulation and control

Abstract

The invention relates to the technical field of energy management, and discloses a multi-energy collaborative zero-carbon park smart energy system based on AI agent autonomous regulation and control. The energy labeling module is used for calculating an energy efficiency index and performing energy efficiency grade labeling on the energy; the demand judgment module is used for predicting an energy demand value of the next regulation and control period; the energy screening module is used for screening target energy from the energy; the intelligent prediction module identifies autonomous regulation and control data through an AI intelligent agent; the autonomous regulation and control module executes an autonomous regulation and control instruction; according to the method, the problem of limitation existing in fixed-dimension and single-angle energy analysis operation can be avoided, a combination mechanism of fluctuation threshold judgment and abnormal value elimination is introduced, the stability of parameter calculation can be improved, errors and misjudgment caused by data fluctuation are avoided, the dynamic screening effect of target energy can be achieved, and the method is suitable for large-scale popularization and application. The overall utilization efficiency of energy is improved, and the efficient and accurate multi-energy cooperative autonomous regulation and control effect is achieved.

Description

AI-based intelligent agent-driven multi-energy collaborative zero-carbon park smart energy system Technical Field

[0001] This invention relates to the field of energy management technology, and more specifically, to a smart energy system for multi-energy collaborative zero-carbon parks based on AI-driven intelligent agents with autonomous control. Background Technology

[0002] With the proposal of dual carbon targets, building zero-carbon parks has become an important path for sustainable development. As the core support for zero-carbon parks, the level of intelligence of smart energy systems directly affects the energy utilization efficiency and carbon emission levels of the parks. Moreover, through multi-dimensional analysis, advanced prediction and AI-driven autonomous control mechanisms, it has become an important way to promote the construction of multi-energy coordinated zero-carbon parks.

[0003] The patent application CN120338458A discloses a multi-energy scheduling and control method and system for a zero-carbon energy system in a smart park. The method includes the following steps: dividing park energy into multiple energy types, including Class I, Class II, Class III, and Class IV energy; periodically predicting the energy demand of each energy demand sub-module within the smart park in the next cycle; determining the energy type allocated to the energy demand sub-module based on a first optimization model; and determining the park energy allocated to the energy demand sub-module based on the energy type using a second optimization model. Existing smart energy systems typically evaluate energy efficiency using a single or fixed dimension, failing to comprehensively analyze the dynamics of the energy supply process from multiple dimensions such as conversion rate, loss rate, peak rate, and margin rate. This results in limitations in the evaluation and analysis process. Furthermore, the lack of a combined processing mechanism for dynamic data fluctuations during multi-dimensional data calculations leads to significant influence from abnormal data on subsequent calculation results, resulting in low accuracy and potential misjudgments in the park's autonomous energy control operations. Consequently, efficient and accurate multi-energy collaborative autonomous control cannot be achieved.

[0004] In view of this, the present invention proposes a multi-energy collaborative zero-carbon smart energy system for industrial parks based on the autonomous regulation of AI agents to solve the above problems. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of existing technologies and achieve the above objectives, this invention provides the following technical solution: a multi-energy collaborative zero-carbon park smart energy system based on AI intelligent agent autonomous control, applied to an energy control platform, comprising: an energy labeling module, which calculates an energy efficiency index through energy efficiency supply parameters, compares the energy efficiency index with preset first and second energy efficiency thresholds respectively, and labels the energy efficiency level as Level 1, Level 2, or Level 3 energy based on the comparison results; and a demand determination module, which collects demand impact data during the control cycle, including production operation data, external environmental data, and production scheduling data. The system uses data to predict energy demand for the next control cycle through an energy demand forecasting model and determines whether to implement autonomous control mode. The energy screening module, under autonomous control mode, calculates the demand difference and maximum supply value based on the next control cycle and selects target energy sources. The intelligent forecasting module aggregates the energy-related parameters and demand difference of the target energy into comprehensive control data and uses an AI agent to identify the autonomous control data of the target energy. The autonomous control module converts the autonomous control data into autonomous control commands, executes these commands through the energy control platform, and allocates the target energy to the park.

[0006] Furthermore, energy efficiency supply parameters include energy conversion efficiency, energy loss rate, peak energy efficiency rate, and energy consumption margin rate. The method for collecting the energy consumption margin rate is as follows: A1: Using a unit of measurement as the standard, query the database to find the duration of energy supply for that unit from the start to the end of the supply, and record it as the energy supply cycle; A2: Starting from the current moment, count backwards along the timeline to mark A consecutive energy supply cycles, calculate the energy supply at the first moment and the energy surplus at the last moment in the A energy supply cycles, divide the energy surplus by the energy supply to calculate the unit margin rate, and sum the A unit margin rates and calculate the average to obtain the energy consumption margin rate. A3: Calculate the median of the margin rate by adding the maximum and minimum values ​​of the unit margin rate and averaging them. Then, calculate the margin fluctuation value by taking the absolute value of the difference between the median and the average margin rate. A4: When the margin fluctuation value is less than or equal to the calibrated fluctuation threshold, record the average margin rate as the energy consumption margin rate. A5: When the margin fluctuation value is greater than the calibrated fluctuation threshold, remove the maximum and minimum values ​​of the unit margin rate and repeat steps A3-A5 until the margin fluctuation value is less than or equal to the calibrated fluctuation threshold. Then, accumulate the remaining unit margin rates and average them to calculate the energy consumption margin rate.

[0007] Furthermore, the energy efficiency index is calculated as follows: Verification events for energy efficiency conversion rate, energy loss rate, peak energy efficiency rate, and surplus consumption rate during historical periods are retrieved from the database one by one, and verification events with correct verification results are recorded as correct events. The confidence level is calculated by dividing the number of correct events by the number of verification events. Based on the principle that the higher the confidence level, the larger the proportional coefficient, the corresponding proportional coefficients are assigned to the energy efficiency conversion rate, energy loss rate, peak energy efficiency rate, and surplus consumption rate in sequence, and then the weighted sum is performed to calculate the energy efficiency index.

[0008] Furthermore, the method for labeling energy efficiency levels is as follows: The energy efficiency index of the energy source is used as the labeling method. Each is compared with the preset first energy efficiency threshold. Second energy efficiency threshold Perform a size comparison, and Greater than ;when Greater than When, the energy is labeled as a Level 1 energy source; when Less than or equal to ,and Greater than When, the energy source is labeled as a Level 2 energy source; when Less than At that time, the energy was labeled as Level 3 energy.

[0009] Furthermore, the method for determining whether to implement the autonomous control mode is as follows: The energy supply and energy consumption values ​​of the park in the current control cycle are retrieved from the database. The energy storage value is calculated by subtracting the energy supply and energy consumption values. The energy supply value of the park in the next control cycle is retrieved. The total energy value is calculated by adding the energy supply value to the energy storage value. When the total energy value is greater than or equal to the energy demand value, it is determined that the autonomous control mode will not be implemented. When the total energy value is less than the energy demand value, it is determined that the autonomous control mode will be implemented.

[0010] Furthermore, the method for calculating the demand difference is as follows: taking the next control cycle as the time base, the predicted energy demand value is subtracted from the total energy value of the park to calculate the theoretical difference value; the energy loss rate of the park within a unit time period is retrieved from the database, the duration of the control cycle is divided by the unit time period to calculate the loss ratio, and the loss ratio is multiplied by the energy loss rate to calculate the energy loss value; the theoretical difference value is added to the energy loss value to calculate the demand difference value.

[0011] Furthermore, the maximum supply value includes the primary supply value, the secondary supply value, and the tertiary supply value. When calculating the primary, secondary, and tertiary supply values, the storage quantities of primary, secondary, and tertiary energy in the next control cycle are calculated sequentially to obtain the primary, secondary, and tertiary storage quantities. The primary, secondary, and tertiary storage quantities are then multiplied by their corresponding energy efficiency conversion rates to calculate the primary, secondary, and tertiary theoretical values. Finally, the primary, secondary, and tertiary theoretical values ​​are subtracted from the preset energy lower limit values ​​to calculate the primary, secondary, and tertiary supply values.

[0012] Furthermore, the selection method for target energy is as follows: when the demand difference is less than or equal to the demand difference, primary energy is recorded as target energy; when the demand difference is greater than the primary supply value, the primary and secondary supply values ​​are added together to obtain the cumulative supply value, and the demand difference is compared with the cumulative supply value; when the demand difference is less than or equal to the cumulative supply value, both primary and secondary energy are recorded as target energy; when the demand difference is greater than the cumulative supply value, primary, secondary, and tertiary energy are all recorded as target energy.

[0013] Furthermore, the energy substance data includes type number and type supply value; the autonomous regulation data includes regulation number, regulation supply value, and regulation sequence number; the training method for the AI ​​agent is as follows: training set construction: collect multiple sets of comprehensive regulation data and autonomous regulation data within a historical time period, bind a set of comprehensive regulation data and autonomous regulation data into a data sample, and divide the data sample and label it as training sample and test sample; model parameters: set the learning rate, number of iterations, and batch size of the deep learning model, and use cross-validation to optimize the deep learning model; training objective: deep learn the autonomous regulation behavior of the target energy in the comprehensive regulation data, and output the autonomous regulation data, upgrading the deep learning model into an AI agent.

[0014] Furthermore, autonomous control instructions include primary control instructions, secondary control instructions, and tertiary control instructions. When converting autonomous control data into autonomous control instructions, firstly, the content of the autonomous control data is parsed, and the content corresponding to primary energy, secondary energy, and tertiary energy is recorded as primary content, secondary content, and tertiary content, respectively. Then, the start time, control authority, and end time are added to the primary content, secondary content, and tertiary content, respectively, to generate primary control instructions, secondary control instructions, and tertiary control instructions.

[0015] The technical effects of the smart energy system for multi-energy collaborative zero-carbon parks based on AI intelligent agent autonomous control are as follows: (1): By analyzing the changes in the conversion, loss, peak and consumption of energy in the energy supply process, this invention can avoid the limitations of fixed-dimensional and single-angle energy analysis operations. In addition, it introduces a combination mechanism of fluctuation threshold judgment and outlier removal in the calculation of energy consumption surplus rate, which can not only improve the stability of parameter calculation and avoid errors and misjudgments caused by data fluctuations, but also achieve the dynamic screening effect of target energy and improve the overall energy utilization efficiency.

[0016] (2): This invention can predict the energy demand of the park in the future through the energy demand prediction model, and realize the effect of advanced prediction of the energy demand of the park. It avoids the problem of lag in autonomous control operation caused by real-time prediction operation. At the same time, combined with AI intelligent agent, it can learn the specific data of autonomous control and provide appropriate autonomous control measures for the dynamic changes of energy in the park at different times. This improves the accuracy and timeliness of autonomous energy control operation in the park, avoids the lag and inaccuracy caused by manual intervention, and realizes the efficient and accurate multi-energy coordinated autonomous control effect. Attached Figure Description

[0017] Figure 1 is a schematic diagram of the module of the multi-energy collaborative zero-carbon park smart energy system based on AI intelligent agent autonomous control provided in Embodiment 1 of the present invention; Figure 2 is a schematic diagram of the process of the multi-energy collaborative zero-carbon park smart energy method based on AI intelligent agent autonomous control provided in Embodiment 2 of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1: Please refer to Figure 1. The multi-energy collaborative zero-carbon park smart energy system based on AI intelligent agent autonomous control described in this example is applied to an energy control platform. It includes: an energy labeling module, which obtains the energy efficiency supply parameters of various energy sources in the park, calculates the energy efficiency index of the energy, and labels the energy with energy efficiency levels, including Level 1, Level 2, and Level 3 energy. In actual production operations, the park contains various types of energy to meet and realize different energy consumption needs. In this example, the energy in the park includes, but is not limited to, natural gas, oil, wind, solar, and thermal energy.

[0020] Energy efficiency supply parameters are parameters used to represent the energy supply process and efficiency of different types of energy in the park for various energy-consuming equipment. They serve as the data basis for identifying and classifying different types of energy. Specifically, energy efficiency supply parameters include energy conversion rate, energy loss rate, peak energy efficiency rate, and surplus energy consumption rate.

[0021] Energy conversion efficiency refers to the ratio of energy that is effectively utilized to the maximum usable energy of the energy source, which can be used to represent the degree of effective energy utilization. The higher the energy conversion efficiency, the higher the energy efficiency index.

[0022] Energy loss rate refers to the ratio of energy that cannot be used for positive work to the energy that is used for positive work, which can be used to express the degree of energy waste. The higher the energy loss rate, the lower the energy efficiency index.

[0023] Peak energy efficiency rate refers to the ratio of the duration during which the energy conversion rate of an energy source is greater than the rated conversion rate to the total duration of energy supply. It can be used to represent the peak duration of energy conversion. The higher the peak energy efficiency rate, the higher the energy efficiency index of the energy source.

[0024] In this embodiment, the energy conversion rate, energy loss rate, and peak energy efficiency rate are all obtained by querying the energy technology parameter table.

[0025] The energy consumption margin ratio is the ratio of the remaining energy after consumption to the total energy before consumption, thus representing the amount of energy consumed. A higher energy consumption margin ratio indicates a lower energy efficiency index. The energy consumption margin ratio is collected as follows: A1: Using a unit of measurement as a standard, retrieve the duration of energy supply from the start to the end of the supply for that unit from a database, recording it as an energy supply cycle; A2: Starting from the current moment, trace back A consecutive energy supply cycles, calculating the energy supply at the first moment and the energy surplus at the last moment of each of the A cycles. Compare the energy surplus with the energy supply to calculate the unit margin ratio, and then sum and average the A unit margin ratios to obtain the mean margin ratio; A3: Add the maximum and minimum unit margin ratios and average them to calculate the median margin ratio. The absolute value of the difference between the median and mean of the margin rate is used to calculate the margin fluctuation value. A4: When the margin fluctuation value is less than or equal to the calibrated fluctuation threshold, the actual energy consumption fluctuates less, and the mean of the margin rate is recorded as the energy consumption margin rate. The calibrated fluctuation threshold refers to the maximum value of the average margin rate recorded as the energy consumption margin rate under normal circumstances, which can limit the fluctuation range of the actual energy consumption in the park and avoid excessive fluctuation. A5: When the margin fluctuation value is greater than the calibrated fluctuation threshold, the actual energy consumption fluctuates more. The maximum and minimum values ​​of the unit margin rate are removed, and steps A3-A5 are repeated until the margin fluctuation value is less than or equal to the calibrated fluctuation threshold. The remaining unit margin rates are then summed and averaged to calculate the energy consumption margin rate.

[0026] It should be noted that by analyzing the conversion, loss, peak and consumption of energy during energy supply, multi-dimensional energy efficiency analysis data can be provided for energy supply to energy-consuming equipment, avoiding the limitations of traditional energy analysis based on characteristic dimensions and single perspectives.

[0027] The energy efficiency index (EI) is a numerical representation of the positive impact of energy supply, providing direct numerical evidence for energy identification and classification. A higher EI indicates higher energy content and greater importance for energy supply. Specifically, the EI is calculated as follows: Verification events for energy conversion efficiency, energy loss rate, peak energy efficiency rate, and surplus energy consumption rate over historical periods are retrieved from a database. Verification events with correct results are recorded as correct events. The number of correct events is compared with the number of verification events to calculate the confidence level. Based on a higher confidence level (and thus a higher proportionality coefficient), corresponding proportionality coefficients are assigned to the energy conversion efficiency, energy loss rate, peak energy efficiency rate, and surplus energy consumption rate, and then weighted and summed to calculate the EI. The formula for calculating the EI is: In the formula, Energy efficiency index, For energy efficiency conversion rate, Energy loss rate, Peak energy efficiency, This refers to the consumption margin rate. , , , These are the proportional coefficients for energy conversion efficiency, energy loss rate, peak energy efficiency, and surplus energy consumption rate, respectively. , , , The sum of is 1, and , , , All are greater than 0.

[0028] Energy efficiency rating is used to represent the efficiency level of energy participating in energy conversion within the park, that is, to classify different energy sources into different energy efficiency ratings; specifically, energy efficiency ratings include Level 1 energy, Level 2 energy, and Level 3 energy; and the importance of Level 1, Level 2, and Level 3 energy is from high to low.

[0029] The method for labeling energy efficiency ratings is as follows: The energy efficiency index of the energy source is used as the labeling method. Each is compared with the preset first energy efficiency threshold. Second energy efficiency threshold Perform a size comparison, and Greater than The preset first and second energy efficiency thresholds are the critical values ​​for the energy efficiency index when energy sources are classified as Level 1 and Level 2, and Level 2 and Level 3, respectively, thus enabling accurate differentiation between Level 1, Level 2, and Level 3 energy sources. Greater than When the energy source has the highest energy efficiency rating and is of the highest importance, it is labeled as a Level 1 energy source; when... Less than or equal to ,and Greater than At this time, if the energy efficiency level and importance of the energy source are both in the middle, then the energy source is labeled as a Level 2 energy source; when Less than At this time, the energy efficiency level and importance level of the energy source are the lowest, so the energy source is labeled as a level three energy source.

[0030] In this embodiment, the energy source is not fixed as a Level 1, Level 2, or Level 3 energy source. It depends on the specific values ​​of the energy efficiency supply parameters of that energy source during a historical period. The number of energy sources corresponding to each energy efficiency level is also not consistent. In order to ensure the complete implementation of the subsequent technical solutions in this embodiment, each energy efficiency level has at least one energy source. For example, Level 1 energy sources are solar energy and wind energy; Level 2 energy sources are natural gas energy and thermal energy; and Level 3 energy sources are petroleum energy.

[0031] The demand determination module collects demand impact data of the park during the control cycle, predicts the energy demand value for the next control cycle using an energy demand forecasting model, and compares the energy demand value with the total energy value to determine whether to implement the autonomous control mode. Demand impact data refers to data that can affect the amount of energy consumed and used by the park within a control cycle. By collecting and analyzing the demand impact data, the energy demand value of the park within a control cycle is predicted and judged. The control cycle is a time period used to analyze and evaluate the energy demand in the park, so that a control cycle can meet the collection of multi-dimensional data and energy demand calculation within the park, while also ensuring that demand impact data can be comprehensively collected within a control cycle. In this embodiment, the duration of the control cycle is not fixed, and it is set according to various factors such as the collection of relevant data, implementation of measures, and equipment operation status in the park during historical periods. For example, the control cycle is 7 days or 15 days.

[0032] Demand impact data includes production operation data, external environmental data, and production scheduling and operation data; production operation data refers to the relevant data on industrial production and living operations carried out in the park during the regulation period, which can provide the most core and critical data for energy demand values; in this embodiment, production operation data includes, but is not limited to, industrial equipment operating time, residential energy consumption, park basic energy consumption, energy consumption time, etc.; production operation data is obtained through database queries.

[0033] External environmental data refers to the relevant data on weather conditions and interference factors affecting the park during the control period, which can provide another dimension of data for energy demand. In this embodiment, external environmental data includes, but is not limited to, temperature distribution curves, humidity change curves, and sunshine duration. External environmental data is acquired through detection by various sensors.

[0034] Production scheduling and operation data refers to the relevant data of the park's planned and prepared operations within the regulation cycle, which can provide another dimension of data for energy demand. In this embodiment, production scheduling and operation data includes, but is not limited to, planned equipment output, production increase ratio, rest day duration, production process type, and new energy consumption. Production scheduling and operation data is obtained through database query.

[0035] The energy demand value is used to describe the amount of energy consumed by the park to meet normal operation and work within a control cycle, so that the energy demand value and the demand impact data are in one-to-one correspondence; in this embodiment, the energy demand value is obtained by statistical analysis of each energy-consuming device.

[0036] The energy demand forecasting model is an artificial intelligence model based on machine learning technology, taking demand impact data as input and energy demand value as output. It enables the energy demand forecasting model to predict the energy demand value of the park in the next regulation cycle based on existing demand impact data, thereby achieving the effect of advanced prediction on the timeline.

[0037] Energy demand forecasting models are not directly obtained; they require prior training. Specifically, the training method for energy demand forecasting models is as follows: Demand impact data and energy demand values ​​for multiple regulation cycles within historical time periods are collected in advance. These multiple sets of demand impact data are then transformed into multiple feature vectors using a sliding window method. Based on the sliding step size, energy demand values ​​are converted into labels corresponding to the demand impact data, with one feature vector corresponding to one label, forming a set of training data. Multiple sets of training data constitute a training set. The demand impact data are arranged according to the collection time, and the prediction time step, sliding step size, and sliding window length are set. The feature vectors are used as input to the energy demand forecasting model, and the energy demand values ​​for the regulation cycle after the prediction time step are used as output. The energy demand values ​​for the next regulation cycle of each training set are used as the prediction target. The training objective is to minimize the sum of prediction errors. This process trains the energy demand forecasting model to generate an energy demand forecasting model that predicts the energy demand values ​​for the next regulation cycle based on the demand impact data.

[0038] Other model parameters in the energy demand forecasting model, such as the target loss value, optimization algorithm, ratio of training set to test set to validation set, and optimization of the loss function, are all obtained through actual engineering implementation and continuous experimental tuning.

[0039] After obtaining the energy demand value for the next control cycle, the energy demand value of the park in the next control cycle can be compared with the controllable value to determine whether to implement the autonomous control mode. Specifically, the method for determining whether to implement the autonomous control mode is as follows: The energy supply value and energy consumption value of the park in the current control cycle are retrieved from the database. The energy storage value is calculated by subtracting the energy supply value and energy consumption value. The energy supply value of the park in the next control cycle is retrieved. The energy supply value and energy storage value are added together to calculate the total energy value. The total energy value is then compared with the energy demand value. When the total energy value is greater than or equal to the energy demand value, the park does not need to implement autonomous control in the next control cycle, and the autonomous control mode is determined not to be implemented. When the total energy value is less than the energy demand value, the park needs to implement autonomous control in the next control cycle, and the autonomous control mode is determined to be implemented.

[0040] The energy screening module, when implementing the autonomous control mode, calculates the demand difference and maximum energy supply of the park based on the next control cycle, and compares the demand difference with the maximum supply to select target energy for allocation to the park. After implementing the autonomous control mode, it is necessary to calculate the demand difference and maximum supply of the park in the next control cycle separately to provide data for autonomous energy control. Specifically, the demand difference refers to the difference between energy demand and energy supply in the park in the next control cycle, serving as a basis for autonomous energy control. The calculation method for the demand difference is as follows: using the next control cycle as the time base, the predicted energy demand is subtracted from the total energy supply of the park to calculate the theoretical difference. The formula for calculating the theoretical difference is: In the formula, This is the theoretical difference value. For the predicted energy demand value, This represents the total energy consumption of the park. The energy loss rate of the park within a unit of time is retrieved from the database. The duration of the control cycle is compared with the unit of time to calculate the loss ratio. This loss ratio is then multiplied by the energy loss rate to calculate the energy loss value. The energy loss rate refers to the proportion of energy lost due to unused consumption within a unit of time, thus improving the accuracy of the demand difference calculation. The formula for calculating the energy loss value is: In the formula, This represents the energy loss value. To regulate the duration of the cycle, For unit duration, The energy loss rate is used as the basis for calculation. The theoretical difference is added to the energy loss value to calculate the demand difference. The formula for calculating the demand difference is: In the formula, This represents the demand difference.

[0041] Maximum supply value refers to the maximum value of energy available for independent external control across all energy efficiency levels within the park, and is used as data for direct comparison with the demand difference value; specifically, maximum supply value includes level 1 supply value, level 2 supply value and level 3 supply value; and corresponds to the maximum supply of level 1 energy, level 2 energy and level 3 energy respectively.

[0042] When calculating the primary, secondary, and tertiary supply values, firstly, the storage quantities of primary, secondary, and tertiary energy in the next control cycle are calculated sequentially to obtain the primary, secondary, and tertiary storage quantities. Then, the primary, secondary, and tertiary storage quantities are multiplied by their corresponding energy efficiency conversion rates to calculate the primary, secondary, and tertiary theoretical values. Finally, the primary, secondary, and tertiary theoretical values ​​are subtracted from the preset energy lower limit values ​​to calculate the primary, secondary, and tertiary supply values.

[0043] It should be noted that the energy lower limit is the minimum value that needs to be retained after autonomous regulation of energy in each regulation cycle, so as to serve as the energy target for emergency rescue in the park; generally speaking, the energy lower limit is 10%-15% of the current storage.

[0044] After obtaining the demand difference and maximum supply value, it is necessary to analyze them under the constraints of the supply balance criterion, and sequentially select target energy sources to be allocated to the park, making these target energy sources the direct objects for subsequent autonomous control by the AI ​​agent. Specifically, the selection method for target energy is as follows: compare the demand difference value with the primary supply value; when the demand difference value is less than or equal to the primary supply value, the primary energy source can meet the park's energy demand in the next control cycle, and is then recorded as the target energy source; when the demand difference value is greater than the primary supply value, the primary energy source cannot meet the park's energy demand in the next control cycle. If the energy supply is sufficient to meet the park's energy demand in the next regulation cycle, the primary and secondary supply values ​​are added together to obtain the cumulative supply value. The demand difference is then compared with the cumulative supply value. When the demand difference is less than or equal to the cumulative supply value, primary and secondary energy can meet the park's energy demand in the next regulation cycle, and both primary and secondary energy are recorded as target energy. When the demand difference is greater than the cumulative supply value, primary and secondary energy cannot meet the park's energy demand in the next regulation cycle, and primary, secondary, and tertiary energy are all recorded as target energy.

[0045] In this embodiment, the sum of primary, secondary, and tertiary energy sources can meet the park's energy needs in the next regulation cycle, and there will be no supply-demand imbalance. Therefore, the specific type of energy corresponding to the target energy source may not be unique.

[0046] The intelligent forecasting module aggregates the energy-related parameters and demand difference of the target energy source into comprehensive control data, and identifies the autonomous control data of the target energy source through an AI agent. After the target energy source is selected, the AI ​​agent can identify and analyze the corresponding autonomous control data based on the energy-related parameters of the target energy source, so that the autonomous control data can serve as a reference for the target energy source to autonomously control the park. Since the specific types and quantities of energy contained in the target energy source are not consistent, the energy-related parameters need to determine the type and quantity of the target energy source. Specifically, the energy-related data includes type number and type supply value. The type number refers to the numerical code corresponding to the specific energy type in the target energy source. Each energy type has a unique type number. For example, the type number of solar energy is TY11, the type number of gas energy is RQ21, and the type number of petroleum energy is SY31.

[0047] Type supply value refers to the maximum energy that a specific energy type in the target energy can supply, which provides an upper limit for the subsequent autonomous regulation of the target energy of that type.

[0048] After obtaining the actual energy data, the actual energy data and the demand difference can be summarized to generate comprehensive regulation data. This comprehensive regulation data can then be input into the AI ​​agent to identify the required autonomous regulation data.

[0049] In this embodiment, the AI ​​agent is an artificial intelligence model based on deep learning technology, which takes comprehensive regulatory data as input data and autonomous regulatory data as output data, enabling it to perform deep learning on comprehensive regulatory data and output the required autonomous regulatory data.

[0050] The autonomous control data is output by the AI ​​agent and is used to meet the comprehensive data requirements for autonomous energy control in the next control cycle of the park. Specifically, the autonomous control data includes control number, control supply value, and control sequence number. The control number refers to the type number of the target energy that needs to be autonomously controlled; the control supply value refers to the supply value of the target energy that needs to be autonomously controlled; the control sequence number refers to the order in which the target energy needs to be autonomously controlled. It should be noted that the control supply value of the target energy is less than or equal to the type supply value; the control sequence number is in ascending order, and the smaller the number, the earlier the autonomous control order.

[0051] Specifically, the training method for the AI ​​agent is as follows: Training set construction: Collect multiple sets of comprehensive regulation data and autonomous regulation data within a historical time period, bind a set of comprehensive regulation data and autonomous regulation data into a data sample, and divide the data sample into training samples and test samples, with a ratio of 3:1 between the number of training samples and test samples; Model parameters: Set the learning rate, number of iterations, and batch size of the deep learning model, and use cross-validation to optimize the deep learning model to avoid overfitting; Training objective: The deep learning model aims to learn the autonomous regulation behavior of the target energy source in the comprehensive regulation data (such as "selection of target energy source → calculation of supply value → priority setting"), and output autonomous regulation data, thus upgrading the deep learning model into an AI agent.

[0052] Once the AI ​​agent is obtained, the comprehensive regulatory data can be input into the AI ​​agent, and the corresponding autonomous regulatory data can be identified.

[0053] The autonomous control module converts autonomous control data into autonomous control commands, executes these commands through the energy control platform, and allocates target energy to the park. After obtaining the autonomous control data, it needs to be converted back into autonomous control commands to provide the energy control platform with allocation instructions for the target energy, ensuring that the target energy is allocated and controlled sequentially and accurately within the park to meet the park's energy needs for the next control cycle. In this embodiment, when converting autonomous control data into autonomous control commands, different energy efficiency levels are used as a benchmark, resulting in three different autonomous control commands. Specifically, the autonomous control commands include Level 1, Level 2, and Level 3 commands.

[0054] When converting autonomous control commands, the content of autonomous control data is first parsed, and the content corresponding to primary energy, secondary energy and tertiary energy is recorded as primary content, secondary content and tertiary content. Then, the start time, control authority and end time are added to the primary content, secondary content and tertiary content respectively, and primary control commands, secondary control commands and tertiary control commands are generated.

[0055] It should be noted that the start time, control authority, and end time are used to represent the start execution time, the authority level at the time of execution, and the end execution time of the autonomous control command, respectively, to ensure the complete and orderly execution of the autonomous control command.

[0056] Upon receiving primary, secondary, and tertiary control commands, the energy control platform can execute these commands, thereby allocating the corresponding target energy to the park in sequence to meet the park's energy needs in the next control cycle. This promotes the park's zero-carbon development, achieves intelligent and autonomous energy control within the park, and prevents future energy supply and demand imbalances.

[0057] Example 2: Please refer to Figure 2. For details not described in this example, please refer to the description in Example 1. This example provides a smart energy method for multi-energy collaborative zero-carbon parks based on AI-powered intelligent agent autonomous control. It is applied to an energy control platform and implemented through a multi-energy collaborative zero-carbon park smart energy system based on AI-powered intelligent agent autonomous control. The method includes: S01: Calculating the energy efficiency index using energy supply parameters; comparing the energy efficiency index with preset first and second energy efficiency thresholds; and labeling the energy efficiency level as Level 1, Level 2, or Level 3 based on the comparison results; S02: Collecting demand impact data during the control cycle and predicting the next energy demand using an energy demand forecasting model. The system calculates the energy demand value for one control cycle and determines whether to implement the autonomous control mode. If the autonomous control mode is implemented, steps S03-S05 are executed. If the autonomous control mode is not implemented, step S02 is repeated. S03: Under the autonomous control mode, the demand difference and maximum supply value are calculated based on the next control cycle, and the target energy is selected from the energy sources. S04: The energy-related parameters and demand difference of the target energy are summarized into comprehensive control data, and the autonomous control data of the target energy is identified by the AI ​​agent. S05: The autonomous control data is converted into autonomous control instructions, and the autonomous control instructions are executed through the energy control platform to allocate the target energy to the park.

[0058] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A multi-energy collaborative zero-carbon smart energy system for industrial parks based on AI-powered autonomous control, applied to an energy control platform, characterized in that: include: The energy labeling module calculates the energy efficiency index based on energy supply parameters, compares the energy efficiency index with preset first and second energy efficiency thresholds, and labels the energy efficiency level as Level 1, Level 2, or Level 3 based on the comparison results. The demand determination module collects demand impact data for the control cycle, including production operation data, external environmental data, and production scheduling data. It predicts the energy demand value for the next control cycle through an energy demand forecasting model and determines whether to implement the autonomous control mode. The energy screening module, under autonomous control mode, calculates the demand difference and maximum supply value based on the next control cycle, and screens out target energy from the energy sources; the intelligent forecasting module summarizes the energy-related parameters and demand difference of the target energy into comprehensive control data, and identifies the autonomous control data of the target energy through an AI agent. The autonomous control module converts autonomous control data into autonomous control commands, which are then executed through the energy control platform to allocate target energy to the park.

2. The multi-energy collaborative zero-carbon smart energy system for industrial parks based on AI intelligent agent autonomous control as described in claim 1, characterized in that, Energy efficiency supply parameters include energy conversion efficiency, energy loss rate, peak energy efficiency rate, and energy consumption margin rate. The method for collecting the energy consumption margin rate is as follows: A1: Using a unit of measurement as the standard, query the database to find the duration of energy supply from the start to the end of the energy supply for a unit of measurement, and record it as the energy supply cycle; A2: Starting from the current moment, count backwards along the timeline to mark A consecutive energy supply cycles, calculate the energy supply at the first moment and the energy margin at the last moment in the A energy supply cycles, divide the energy margin by the energy supply to calculate the unit margin rate, and sum the A unit margin rates and calculate the average margin rate. A3: Add the maximum value of the unit margin rate to the minimum value of the unit margin rate and then average them to calculate the median value of the margin rate. Then, take the absolute value of the difference between the median value of the margin rate and the mean value of the margin rate to calculate the margin fluctuation value. A4: When the margin fluctuation value is less than or equal to the calibrated fluctuation threshold, the average margin rate is recorded as the energy consumption margin rate; A5: When the margin fluctuation value is greater than the calibrated fluctuation threshold, the maximum and minimum values ​​of the unit margin rate are removed, and steps A3-A5 are repeated until the margin fluctuation value is less than or equal to the calibrated fluctuation threshold. Then, the remaining unit margin rates are accumulated and averaged to calculate the energy consumption margin rate.

3. The multi-energy collaborative zero-carbon smart energy system for industrial parks based on AI intelligent agent autonomous control as described in claim 2, characterized in that, The energy efficiency index is calculated as follows: Verification events for energy efficiency conversion rate, energy loss rate, peak energy efficiency rate, and surplus energy consumption rate over historical periods are retrieved from the database one by one, and verification events with correct verification results are recorded as correct events. The confidence level is calculated by dividing the number of correct events by the number of verification events. Based on the principle that the higher the confidence level, the higher the proportional coefficient, the corresponding proportional coefficients are assigned to the energy efficiency conversion rate, energy loss rate, peak energy efficiency rate, and surplus energy consumption rate, and then the weighted sum is calculated to obtain the energy efficiency index.

4. The multi-energy collaborative zero-carbon smart energy system for industrial parks based on AI intelligent agent autonomous control as described in claim 3, characterized in that, The method for labeling energy efficiency ratings is as follows: The energy efficiency index of the energy source is used as the labeling method. Each is compared with the preset first energy efficiency threshold. Second energy efficiency threshold Perform a size comparison, and Greater than ;when Greater than When, the energy is labeled as a Level 1 energy source; when Less than or equal to ,and Greater than At that time, the energy source was labeled as a Level 2 energy source; when Less than At that time, the energy was labeled as Level 3 energy.

5. The multi-energy collaborative zero-carbon smart energy system for industrial parks based on AI intelligent agent autonomous control as described in claim 4, characterized in that, The method for determining whether to implement the autonomous control mode is as follows: query the energy supply and energy consumption values ​​of the park in the current control cycle through the database, calculate the energy storage value by subtracting the energy supply and energy consumption values; Find out the energy supply value of the park in the next regulation cycle, add the energy supply value to the energy storage value and calculate the total energy value; When the total energy consumption is greater than or equal to the energy demand, the autonomous control mode is not executed; when the total energy consumption is less than the energy demand, the autonomous control mode is executed.

6. The multi-energy collaborative zero-carbon smart energy system for industrial parks based on AI intelligent agent autonomous control as described in claim 5, characterized in that, The method for calculating the demand difference is as follows: taking the next control cycle as the time base, the predicted energy demand value is subtracted from the total energy value of the park to calculate the theoretical difference value; the energy loss rate of the park within a unit time is retrieved from the database, the duration of the control cycle is divided by the unit time to calculate the loss ratio, and the loss ratio is multiplied by the energy loss rate to calculate the energy loss value; the theoretical difference value is added to the energy loss value to calculate the demand difference value.

7. The multi-energy collaborative zero-carbon smart energy system for industrial parks based on AI intelligent agent autonomous control as described in claim 6, characterized in that, The maximum supply value includes primary supply value, secondary supply value, and tertiary supply value; When calculating the primary, secondary, and tertiary supply values, the storage quantities of primary, secondary, and tertiary energy in the next control cycle are calculated sequentially to obtain the primary, secondary, and tertiary storage quantities. The primary, secondary, and tertiary storage quantities are then multiplied by their corresponding energy efficiency conversion rates to calculate the primary, secondary, and tertiary theoretical values. Finally, the primary, secondary, and tertiary theoretical values ​​are subtracted from the preset energy lower limit values ​​to calculate the primary, secondary, and tertiary supply values.

8. The multi-energy collaborative zero-carbon smart energy system for industrial parks based on AI intelligent agent autonomous control as described in claim 7, characterized in that, The selection method for target energy is as follows: when the demand difference is less than or equal to the demand difference, the primary energy source is recorded as the target energy source; When the demand difference is greater than the primary supply value, the primary supply value and the secondary supply value are added together to obtain the cumulative supply value, and the demand difference value is compared with the cumulative supply value. When the demand difference is less than or equal to the cumulative supply, both primary and secondary energy sources are recorded as target energy sources. When the demand difference exceeds the cumulative supply, primary energy, secondary energy, and tertiary energy are all recorded as target energy.

9. The multi-energy collaborative zero-carbon smart energy system for industrial parks based on AI intelligent agent autonomous control as described in claim 8, characterized in that, Energy-related data includes type number and type supply value; autonomous regulation data includes regulation number, regulation supply value, and regulation sequence number; the training method for the AI ​​agent is as follows: training set construction: collect multiple sets of comprehensive regulation data and autonomous regulation data within a historical time period, bind a set of comprehensive regulation data and autonomous regulation data into a data sample, and divide and label the data sample as training samples and test samples; model parameters: set the learning rate, number of iterations, and batch size of the deep learning model, and use cross-validation to optimize the deep learning model; Training objective: To enable deep learning to regulate the autonomous regulation behavior of target energy in comprehensive data and output autonomous regulation data, thereby upgrading the deep learning model into an AI agent.

10. The multi-energy collaborative zero-carbon smart energy system for industrial parks based on AI intelligent agent autonomous control as described in claim 9, characterized in that, Autonomous control instructions include primary control instructions, secondary control instructions, and tertiary control instructions. When converting autonomous control data into autonomous control instructions, firstly, the content of the autonomous control data is parsed, and the content corresponding to primary energy, secondary energy, and tertiary energy is recorded as primary content, secondary content, and tertiary content, respectively. Then, the start time, control authority, and end time are added to the primary content, secondary content, and tertiary content, respectively, to generate primary control instructions, secondary control instructions, and tertiary control instructions.

Images