Power line carrier energy-carbon linkage control method and system based on hybrid prediction

Abstract

The invention discloses a hybrid prediction-based power carrier energy-carbon linkage control method and system, and the method comprises the steps: obtaining power utilization data, weather data, policy data and economic data, inputting the data into a hybrid prediction model, and predicting a power demand prediction value in a future time period; a multi-target constraint optimization model is constructed, and an energy-carbon linkage distribution strategy in the future time period is solved and obtained, including the electric quantity purchase amount, the clean energy power supply amount, the energy storage charge and discharge amount and the energy storage charge state in the future time period; and according to an energy-carbon linkage distribution strategy in a future time period, correspondingly generating an electric power purchase instruction, an electric power generation instruction and an electric power storage instruction, and correspondingly sending the electric power purchase instruction, the electric power generation instruction and the electric power storage instruction to an electric power purchase client, an electric power generation device and an electric power storage device in an electric power carrier communication mode to control electric power purchase, generation and storage. According to the invention, the method can achieve the precise prediction of the power demand in a complex environment, dynamically adjusts the energy supply according to the prediction result, and effectively controls the carbon emission.

Description

Technical Field

[0002] This application relates to the technical field of power energy management and distribution, specifically to a power carrier energy carbon linkage control method and system based on hybrid prediction. Background Technology

[0003] Significant progress has been made in the field of smart grids and energy management in recent years, playing a vital role in promoting the rational use of energy. However, existing power load forecasting and energy allocation strategies mostly rely on historical data statistical analysis, lacking the ability to efficiently process real-time data and accurately predict future trends. With increasingly complex grid structures and constantly changing user demands, especially in the context of a low-carbon economy, how to achieve efficient energy use and effective carbon emission control has become a key concern for the industry.

[0004] Currently, the common approach to solving electricity demand forecasting and energy allocation problems relies on static models and fixed algorithms. These methods build models based on existing historical data, predict electricity demand through statistical analysis, and then allocate energy accordingly. While this traditional approach improves system efficiency to some extent and provides a basic basis for energy management, it also has significant drawbacks. For example, due to the lack of effective time-series forecasting models and dynamic adjustment strategies, the prediction accuracy is low when facing complex environmental changes, such as sudden weather changes, electricity price fluctuations, and subsidy policies. This makes it impossible to respond to electricity demand and optimize energy allocation in a timely and accurate manner, leading to energy waste, ineffective carbon emission control, and a decline in user experience. Furthermore, traditional methods cannot achieve accurate electricity demand forecasting and lack flexible linkage control mechanisms, failing to adjust energy allocation strategies in real time based on forecast results, thus making it difficult to meet low-carbon and environmental protection goals. Summary of the Invention

[0005] In order to achieve accurate prediction of electricity demand in complex environments and dynamically adjust energy supply based on the prediction results to effectively control carbon emissions, this application provides a power carrier energy carbon linkage control method and system based on hybrid prediction.

[0006] In a first aspect, this application provides a power line carrier energy-carbon linkage control method based on hybrid prediction, comprising: To acquire electricity consumption data, weather data, policy data, and economic data transmitted via power line carrier communication; The acquired electricity consumption data, weather data, policy data, and economic data are input into a hybrid forecasting model to predict electricity demand for future periods. The hybrid forecasting model adopts a hierarchical forecasting framework, including a basic decision layer and a decision fusion layer. The basic forecasting layer is used in advance to obtain the initial electricity demand forecast and its probability distribution. The basic forecasting layer is equipped with a time series model layer based on LSTM, a causal model layer based on XGBoost, and an uncertainty model layer based on a Bayesian model. The decision fusion layer is then used to weightedly integrate and obtain the electricity demand forecast and its probability distribution. A multi-objective constrained optimization model is constructed to solve for the future energy-carbon linkage allocation strategy, including the future electricity purchase amount, clean energy supply amount, energy storage charging and discharging amount, and energy storage state of charge. The multi-objective constrained optimization model takes minimizing total carbon emissions and electricity costs as the objective function, and sets the following constraints: the future electricity purchase amount, clean energy supply amount, energy storage charging and discharging amount and the predicted value of electricity demand in the future period meet the supply and demand balance; the future clean energy supply amount does not exceed the upper limit of clean energy supply; the future electricity purchase amount does not exceed the upper limit of grid electricity purchase; and the future energy storage capacity does not exceed the preset energy storage capacity. According to the future energy-carbon linkage allocation strategy, corresponding power purchase instructions, power generation instructions, and power storage and release instructions are generated and sent to the power purchase client, power production device, and power storage and release device via power carrier communication to control power purchase, production, and storage and release.

[0007] By adopting the above scheme, a hybrid forecasting model with a hierarchical forecasting framework is used to make high-precision forecasts of electricity demand. A multi-objective constrained optimization model is constructed, and an energy-carbon linkage control strategy is introduced. Energy supply is dynamically adjusted according to the electricity demand forecast results to effectively control carbon emissions. Power line carrier communication technology is used to complete the timely control of real-time electricity purchase, production capacity and storage, and timely response to changes in real-time electricity demand.

[0008] Preferred options also include: A dynamic adjustment model for predictive time periods is constructed and trained. This model employs a deep learning algorithm. The input to the model consists of real-time acquired electricity consumption area information and seasonal information transmitted via power line carrier communication, as well as key electricity consumption data and key weather data extracted based on these information. The output is the length of the predicted time period. The model is trained using historically acquired electricity consumption area information and seasonal information, key electricity consumption data and key weather data extracted based on these information, and a preset time length range within which the predicted and actual electricity demand values ​​for future periods have an error less than a preset error value. The key electricity consumption data and key weather data extracted based on the acquired electricity consumption area information and seasonal information are extracted according to rules that preset key electricity consumption characteristics for different electricity consumption areas and preset key weather information for different area and season information. The real-time acquired electricity consumption area information and seasonal information, as well as key electricity consumption data and key weather data extracted based on the real-time acquired electricity consumption area information and seasonal information, are input into the dynamic adjustment model for the forecast period to obtain the forecast period length; the future period length in the forecast value of electricity demand for the future period is adjusted according to the forecast period length.

[0009] By adopting the above scheme, a dynamic adjustment model for forecast periods is constructed and trained. Based on real-time electricity consumption area information, seasonal information, and corresponding key electricity consumption data and key weather data, the length of the forecast period is obtained. The future length of the electricity demand forecast is adjusted to achieve high-precision forecasting of electricity demand. The forecast period is flexibly adjusted according to the actual situation, so that energy allocation is more in line with actual needs, better copes with complex environmental changes, and reduces carbon emissions.

[0010] Preferably, it further includes: selecting to obtain the electricity demand forecast value and the probability distribution of the electricity demand forecast value by dynamically weighted integration of the set decision fusion layer instead of obtaining the electricity demand forecast value and the probability distribution of the electricity demand forecast value by weighted integration of the set decision fusion layer; including: A decay mechanism is designed for the time-series model layer, including: defining a time decay function, with the formula: α t =e -λ·t In the formula, t is the prediction time step, and λ is the decay coefficient, which is obtained by fitting historical errors; the defined time decay function is integrated into the set time series model basic weights; A volatility response mechanism is designed for the causal model layer, and a volatility index is defined with the following formula: In the formula, i represents the fluctuation factor, and the quantity is n, which comes from weather data, policy data, and economic data; Let i be the rolling standard deviation of the current factor i. Define the historical average volatility level of current factor i; integrate the defined volatility index into the set causal model base weights; For the uncertain model layer, a risk triggering mechanism is designed, and a multidimensional risk index is defined with the following formula: γ=w1·R 外部因素异常度 +w2·R 模型预测不确定性 +w1·R 突发事件触发度 In the formula, w1, w2, and w3 are the weights of each indicator; R 外部因素异常度 R is obtained by weighting the values ​​after standardizing numerical anomalies based on weather data, policy data, and economic data; 模型预测不确定性 R is obtained by standardizing and weighting the prediction interval width and model divergence of the mixture model based on historical adjacent time periods and without optimization of the mixture model parameters; 突发事件触发度 This is a quantitative value used to determine whether historical emergencies are triggered based on real-time weather data, policy data, and economic data; when γ is greater than a preset threshold, the calculated multidimensional risk index is integrated into the set causal model base weights. The dynamically adjusted weights of the time series model layer, causal model layer, and uncertain model layer are normalized, and the corresponding weights are weighted and integrated to obtain the electricity demand forecast and the probability distribution of the electricity demand forecast.

[0011] By adopting the above scheme, a time-series model layer decay mechanism, a causal model layer fluctuation response mechanism, and an uncertain model layer risk triggering mechanism are designed. The weights of each model layer are dynamically adjusted and normalized before being weighted and integrated to obtain the power demand forecast and probability distribution, thereby further improving the robustness and accuracy of the forecast and better coping with complex environmental changes.

[0012] Preferred options also include: Obtain the actual value of electricity demand for future periods, and use the actual value of electricity demand and the corresponding electricity consumption data, weather data, policy data and economic data as the basis for incremental training data; A hierarchical dynamic optimization strategy is designed to perform incremental training for different types of model layers according to the matching hierarchical dynamic optimization strategy to complete the model parameter optimization. Specifically, the hierarchical dynamic optimization strategy matching the temporal model layer adopts a combination of real-time incremental training and periodic incremental training, the hierarchical dynamic optimization strategy matching the causal model layer adopts a combination of triggered incremental training and periodic incremental training, and the hierarchical dynamic optimization strategy matching the uncertain model layer adopts a combination of periodic incremental training and triggered training.

[0013] By adopting the above scheme, the actual value of electricity demand in future periods is obtained as the basis for incremental training data. The parameters are optimized by incremental training of different types of model layers according to the hierarchical dynamic optimization strategy, and the prediction model is continuously optimized to improve the robustness and accuracy of the prediction.

[0014] Preferred options also include: The current guidance scenario type is determined based on the obtained policy and economic data, including: if the current carbon trading price is greater than the preset carbon trading price, it is identified as a strong emission reduction guidance scenario; if the current electricity price is less than the preset electricity price, it is identified as a cost-priority guidance scenario; if both the current carbon trading price and the current electricity price are greater than or less than the preset electricity price, it is identified as a strong emission reduction guidance scenario; if neither the current carbon trading price nor the current electricity price is greater than or less than the preset electricity price, it is identified as a balanced optimization guidance scenario. Based on the current guidance scenario type, the weight settings of total carbon emissions and electricity costs in the objective function are dynamically adjusted. Specifically, for the strong emission reduction guidance scenario type, the weight range of total carbon emissions should be set to be greater than the weight range of electricity costs; for the cost priority guidance scenario type, the weight range of total carbon emissions should be set to be less than the weight range of electricity costs; and for the balanced optimization guidance scenario type, the maximum weight difference between the weight range of total carbon emissions and the weight range of electricity costs should be less than the preset weight difference.

[0015] By adopting the above scheme, the current guidance scenario type is determined in advance, and then the weight settings of total carbon emissions and electricity costs in the objective function are dynamically adjusted according to different guidance scenario types. This better adapts to different market environments and policy requirements, achieves a flexible balance between energy allocation strategies and energy conservation and emission reduction and cost control, and meets the dual goals of low-carbon environmental protection and maximizing economic benefits.

[0016] Preferred options also include: Considering the setting of carbon emission quotas, the carbon emission quotas for users in future periods are obtained, and a modified objective function is selected to replace the original objective function. A multi-objective constrained optimization model is then constructed to solve for the energy-carbon linkage allocation strategy in future periods. The modified objective function aims to minimize the carbon quota penalty and the normalized electricity cost, and the formula is: minα·Cost total '+β·Carbon penalty In the formula, Cost total ’ Cost represents the normalized cost of electricity. total Cost of electricity minCost is the lowest historical cost for users. max The user's highest historical cost; Carbon penalty Carbon as a penalty for carbon quotas cap γ represents the user's carbon emission allowance for the future period; γ is the penalty coefficient; α and β are the weights of the corresponding terms.

[0017] By adopting the above scheme, taking carbon emission quotas into account, a multi-objective constrained optimization model is constructed by replacing the original objective function with the modified objective function to solve the energy-carbon linkage optimization allocation strategy, thereby achieving high-precision prediction of electricity demand, reducing supply-demand imbalances, and reducing energy waste.

[0018] Preferred options also include: Before the generated power purchase instruction, power generation instruction, and generated power storage instruction are sent to the power purchase client, power production unit, and power storage unit respectively via power line carrier communication, each instruction is pre-encoded. Upon receiving the instruction verification results from the power purchase client, power production unit, and power storage unit, and the result of instruction verification failure, retransmission is selected based on the result of instruction verification failure.

[0019] By adopting the above scheme, the instructions are encoded before being sent, and instructions that fail verification are retransmitted based on the verification results of the receiver, thus ensuring the accuracy and reliability of the transmission of energy-carbon linkage control instructions.

[0020] Secondly, this application provides a power line carrier energy-carbon linkage control system based on hybrid prediction, comprising: The power data acquisition module is used to acquire electricity consumption data, weather data, policy data, and economic data transmitted via power line carrier communication. The electricity demand forecasting module is used to input acquired electricity consumption data, weather data, policy data, and economic data into a hybrid forecasting model to predict electricity demand for future periods. The hybrid forecasting model adopts a hierarchical forecasting framework, including a basic decision layer and a decision fusion layer. The basic forecasting layer is used to obtain the initial electricity demand forecast and its probability distribution in advance. The basic forecasting layer is equipped with a time series model layer based on LSTM, a causal model layer based on XGBoost, and an uncertainty model layer based on a Bayesian model. The decision fusion layer is then used to weightedly integrate and obtain the electricity demand forecast and its probability distribution. The energy and carbon allocation strategy acquisition module is used to construct a multi-objective constrained optimization model to solve for and obtain the energy and carbon linkage allocation strategy for future periods, including the amount of electricity purchased, the amount of clean energy supplied, the amount of energy storage charged and discharged, and the state of charge of energy storage for future periods. The multi-objective constrained optimization model takes minimizing total carbon emissions and electricity costs as its objective function, and sets the following constraints: the amount of electricity purchased, the amount of clean energy supplied, the amount of energy storage charged and discharged for future periods must be in supply and demand balance with the predicted value of electricity demand for future periods; the amount of clean energy supplied for future periods must not exceed the upper limit of clean energy supply; the amount of electricity purchased for future periods must not exceed the upper limit of grid electricity purchase; and the energy storage capacity for future periods must not exceed the preset energy storage capacity. The energy and carbon allocation strategy execution module is used to generate electricity purchase instructions, electricity generation instructions, and electricity storage and release instructions according to the energy and carbon linkage allocation strategy for future periods. These instructions are then sent to the electricity purchase client, the electricity production capacity device, and the electricity storage and release device via power line carrier communication to control electricity purchase, production capacity, and storage and release.

[0021] By adopting the above scheme, a hybrid forecasting model is used to make high-precision predictions of electricity demand, reducing supply and demand imbalances and energy waste; an energy-carbon linkage optimization allocation strategy is obtained based on a multi-objective constrained optimization model to effectively control carbon emissions and improve environmental performance; and power line carrier communication is used to send commands to control electricity purchase, production capacity and storage, so as to achieve rational energy allocation.

[0022] Thirdly, this application provides a computer-readable storage medium including a stored computer program, wherein the computer program, when running, controls the device where the computer-readable storage medium is located to perform the method described above.

[0023] Fourthly, this application provides a computer device, the computer device including a memory, a processor and a program stored in the memory and executable thereon, the program being executed by the processor to implement the steps of the method described above.

[0024] In summary, this application has the following beneficial effects: 1. A hybrid forecasting model using a hierarchical forecasting framework is used to accurately predict electricity demand, reducing supply-demand imbalances and energy waste; a multi-objective constrained optimization model is constructed, and an energy-carbon linkage control strategy is introduced to dynamically adjust energy supply based on electricity demand forecasting results, effectively controlling carbon emissions; power line carrier communication technology is used to achieve timely control of real-time electricity purchase, production capacity, and storage, enabling faster response to real-time changes in electricity demand. 2. The forecast period length is adjusted by dynamically adjusting the forecast period model, and multiple dynamic adjustment mechanisms are used to make the hybrid forecast model adapt to different situations; the model parameters are optimized by incremental training through a hierarchical dynamic optimization strategy, and the weight of the objective function is adjusted according to different guidance scenario types to improve the accuracy of power demand forecasting. 3. Consider the objective function of carbon emission quota transformation to further optimize energy and carbon allocation; encode and verify the retransmission of instructions to ensure the reliability of instruction transmission, and ultimately achieve effective control over electricity purchase, production capacity and storage. Attached Figure Description

[0025] Figure 1 This is a flowchart of the power carrier energy-carbon linkage control method based on hybrid prediction described in a specific embodiment; Figure 2 This is a schematic diagram of the power carrier energy-carbon linkage control system based on hybrid prediction described in a specific embodiment. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0027] like Figure 1 As shown in the figure, this application discloses a power carrier energy-carbon linkage control method based on hybrid prediction, including steps such as data acquisition, power demand prediction, energy-carbon allocation strategy solution and strategy execution, to achieve the effects of accurate prediction of power demand, optimization of energy allocation and control of carbon emissions; the following is a further detailed description of this application.

[0028] S1. Acquire electricity consumption data, weather data, policy data, and economic data transmitted via power line carrier communication.

[0029] Specifically, to achieve electricity demand forecasting, it is necessary to acquire electricity-related data, such as user electricity consumption data. A sampling time window can be set and data can be sampled in real time. To ensure the accuracy of electricity demand forecasting, factors affecting user electricity consumption should be considered. Weather data, policy data, and economic data should be collected simultaneously with user electricity consumption data.

[0030] To ensure the accuracy of the acquired data, this embodiment employs power line carrier communication technology to acquire electricity consumption data, weather data, policy data, and economic data transmitted via power line carrier communication. Specifically, electricity consumption data is typically collected and transmitted to the data management platform via devices such as electricity meters and PLCs installed at various electricity consumption points. The data management platform synchronizes with the meteorological department's backend server to acquire meteorological data, such as temperature, humidity, and light intensity, collected by locally installed meteorological monitoring equipment and stored on the meteorological department's backend server. The data management platform also synchronizes with the backend servers of government agencies and financial institutions such as power companies to acquire energy-related policy documents issued by the government, such as carbon emission standards and energy subsidy policies; and acquires economic data, such as electricity price fluctuations and carbon trading prices, released by financial institutions and statistical departments.

[0031] S2. Input the acquired electricity consumption data, weather data, policy data, and economic data into the hybrid forecasting model to predict the electricity demand forecast and the probability distribution of the electricity demand forecast for future periods.

[0032] To ensure accurate electricity demand forecasting under various unforeseen circumstances, a hybrid forecasting model is adopted. Specifically, the hybrid forecasting model employs a hierarchical forecasting framework, including a basic forecasting layer and a decision fusion layer with multiple forecasting models. In this embodiment, the basic forecasting layer specifically includes a time-series model layer based on LSTM, a causal model layer based on XGBoost, and an uncertainty model layer based on a Bayesian model. The initial electricity demand forecast and its probability distribution are obtained from the basic forecasting layer containing multiple forecasting models. Then, the decision fusion layer uses weighted integration to obtain the electricity demand forecast and its probability distribution. The following details the process of obtaining the electricity demand forecast and its probability distribution using the specific hybrid forecasting model.

[0033] First, for the time series model layer built on LSTM, the time series model layer is designed with an LSTM network structure, including the number of LSTM layers, the number of units in each layer, and dropout layers. The input layer of the model layer takes a data sequence containing currently collected electricity consumption data, weather data, policy data, and economic data as input. The output layer of the model is a fully connected layer used to output the predicted value of electricity demand. To obtain the probability distribution of the predicted value, additional nodes are added to the output layer to output quantiles, thus obtaining the predicted probability distribution. This is obtained by inputting preprocessed historical data sequences (electricity consumption data, weather data, policy data, and economic data) into the LSTM model for training. The data sequence containing currently collected electricity consumption data, weather data, policy data, and economic data is input into the trained LSTM model to obtain the initial predicted value and quantile of electricity demand, and the corresponding predicted probability distribution is calculated.

[0034] For the causal model layer built on XGBoost, which automatically learns the relationship between external causal variables (weather data, policy data, and economic data) and the target variable (electricity demand forecast), the input layer of the model takes a data sequence containing electricity consumption data, weather data, policy data, and economic data as input, and the output layer outputs the initial electricity demand forecast, which is obtained through training with historical data sequences. The initial electricity demand forecast is obtained by inputting a data sequence containing currently collected electricity consumption data, weather data, policy data, and economic data into the causal model layer. Considering that the causal model layer cannot directly obtain the probability distribution of the electricity demand forecast, multiple predictions can be made. Each prediction introduces a small random perturbation, such as a small random scaling or translation of the features, to obtain multiple prediction values. The probability distribution of the initial electricity demand forecast is calculated based on the distribution of these prediction values.

[0035] For the uncertain model layer constructed based on the Bayesian model, the input features of the input layer of the uncertain model layer based on the Bayesian model include deterministic variables and uncertain variables. The deterministic variables are data sequences containing electricity consumption data, weather data, policy data, and economic data. Uncertain variables need to be extracted from the collected weather data, policy data, and economic data through data analysis, such as: weather forecast errors (temperature prediction errors, wind speed prediction errors), policy change risk (probability of temporary electricity price adjustments), and lagging economic indicators (fluctuations in the electricity consumption elasticity coefficient). The model construction package... This includes: selecting a prior distribution based on historical input data, such as temperature error and policy change probability; defining a likelihood function based on the input data to describe the relationship between the input data and the electricity demand value; calculating the posterior distribution of the electricity demand value using Bayes' theorem, combined with the prior distribution and the likelihood function; obtaining the predicted electricity demand value and its probability distribution for a future period based on the posterior distribution; and inputting a data sequence containing currently collected electricity consumption data, weather data, policy data, and economic data into the uncertain model layer to obtain the initial predicted electricity demand value and its fractional value, and then calculating the corresponding predicted probability distribution.

[0036] Secondly, the power demand forecast and its probability distribution are obtained by weighted integration using the established decision fusion layer. The weights between the power demand forecast and its probability distribution obtained from different model layers can be determined based on historical forecast data or expert experience.

[0037] S3. Construct a multi-objective constrained optimization model and solve it to obtain the energy-carbon linkage allocation strategy for future periods.

[0038] Specifically, the decision variables of the multi-objective constrained optimization model are defined as: the energy-carbon linkage allocation strategy for future time period (t), specifically including: the amount of electricity purchased in future time period t. Clean energy power supply Energy storage charging and discharging capacity and and the state of charge (SOC) of energy storage t .

[0039] With the objective function of minimizing total carbon emissions and electricity costs, the formula is: minα·Cost total +β·Carbon total Where α and β are the weights of the corresponding indicator items, and their sum is 1; Total cost: In the formula, π represents the time-of-use electricity price for future time period t. clean c represents the clean energy operation and maintenance cost for future time period t. ch c dis The energy storage charging / discharging loss cost for future time period t; Total emissions: In the formula, λ is the carbon emission factor of the power grid. clean It is a carbon emission factor for clean energy.

[0040] Setting constraints includes: Supply and demand need to be considered. The future electricity purchase volume, clean energy supply volume, and energy storage charging / discharging volume should be set to balance the future electricity demand forecast. The formula is: Where D t The above steps provide the predicted electricity demand values ​​for the future time period t.

[0041] The future clean energy power supply will not exceed the clean energy power supply ceiling, as shown in the formula: In the formula, This represents the upper limit of actual clean energy output under weather conditions within a future time period t, which can be determined based on the upper limit of actual clean energy output under the same weather conditions and time periods in the past. The amount of electricity purchased in the future period shall not exceed the grid's power purchase limit, as shown in the formula: In the formula, This represents the upper limit for electricity purchases in the future time period t; The energy storage capacity in the future period will not exceed the preset energy storage capacity, as shown in the formula: SOC min ≤SOC t ≤SOC max Where, SOC min SOC max These are the preset maximum energy storage capacity and the preset minimum energy storage capacity, respectively. In addition, charging and discharging are mutually exclusive, meaning that only charging or discharging can occur at the same time.

[0042] A genetic algorithm is used to solve for the energy and carbon linkage allocation strategy in the future time period, and the optimal energy and carbon linkage allocation strategy is obtained.

[0043] S4. Implement the future time period energy-carbon linkage allocation strategy.

[0044] Specifically, based on the future energy-carbon linkage optimization allocation strategy, corresponding electricity purchase instructions, electricity generation instructions, and electricity storage and release instructions are generated. These instructions are then sent to electricity purchase clients, power generation devices, and electricity storage and release devices via power line carrier communication to control electricity purchase, generation, and storage and release. Electricity purchase instructions are sent to electricity purchase clients to guide them in purchasing the appropriate quantity of electricity; electricity generation instructions are sent to power generation devices, such as power plants and solar power plants, to ensure they produce electricity according to plan; and electricity storage and release instructions are sent to electricity storage and release devices, such as battery energy storage systems, to control their charging or discharging operations.

[0045] Furthermore, to ensure the error-free execution of the energy and carbon allocation strategy, before the generated electricity purchase instructions, electricity generation instructions, and generated electricity storage instructions are sent to the electricity purchase client, electricity production unit, and electricity storage unit respectively via power line carrier communication, each instruction is pre-encoded. Upon receiving the instruction verification results (pass or fail) from the electricity purchase client, electricity production unit, and electricity storage unit, retransmission is selected based on the failure result. Different instructions can adopt different encoding forms, and instructions containing CRC check codes are sent to each device to determine the accuracy of the current instruction transmission through CRC check.

[0046] In a specific embodiment, considering the varying impacts of different electricity consumption areas and seasons on electricity demand, a dynamic adjustment model for forecast periods is constructed. Based on real-time regional and seasonal information, the length of the forecast period is dynamically adjusted, thereby improving the accuracy of electricity demand forecasting, further optimizing the energy-carbon linkage allocation strategy, better meeting electricity demand under different circumstances, and reducing energy waste and carbon emissions. The method also includes: A dynamic adjustment model for predicting electricity demand is constructed and trained. Regional characteristics, seasonal characteristics, and weather characteristics are factors that significantly influence electricity demand. The dynamic adjustment model employs a deep learning algorithm. The model's input consists of real-time electricity consumption area information and seasonal information transmitted via power line carrier communication, as well as key electricity consumption data and key weather data extracted based on these information. The output is the length of the predicted period. During model training, historically acquired electricity consumption area and seasonal information, key electricity consumption data and key weather data extracted based on these information, and a preset time range within which the predicted electricity demand for future periods has an error less than a preset error value are used as training data to generate the dynamic adjustment model for the predicted period. The key electricity consumption data and key weather data extracted based on the acquired electricity consumption area information and seasonal information are extracted according to the rules of preset key electricity consumption characteristics for different electricity consumption areas and preset key weather information for different area information and seasonal information. In this embodiment, the types of electricity consumption areas are divided into industrial areas, commercial areas and residential areas. The key electricity consumption data extracted for industrial areas is the load fluctuation coefficient, i.e. (maximum load of the time period - minimum load of the time period) / average load of the time period; the key electricity consumption data extracted for commercial areas is the peak-valley difference rate, i.e. (peak electricity consumption of the time period - valley electricity consumption of the time period) / peak electricity consumption of the time period; and the key electricity consumption data extracted for residential areas is the load factor, i.e. average load of the time period / maximum load of the time period. The seasonal characteristics are divided into spring, summer, autumn and winter. The key data extracted for each season are temperature, humidity and light intensity, but the proportion of key data extracted in different seasons is different.

[0047] The real-time acquired electricity consumption area information and seasonal information, as well as key electricity consumption data and key weather data extracted based on the real-time acquired electricity consumption area information and seasonal information, are input into the dynamic adjustment model for the forecast period to obtain the forecast period length; the future period length in the forecast value of electricity demand for the future period is adjusted according to the forecast period length.

[0048] In a specific embodiment, by designing a time-series model layer attenuation mechanism, a causal model layer fluctuation response mechanism, and an uncertain model layer risk triggering mechanism, the weights of each model layer are dynamically adjusted and normalized before weighted integration to obtain the predicted electricity demand value and probability distribution, thereby further improving the robustness and accuracy of the prediction to better cope with complex environmental changes; the method includes: Instead of using the set decision fusion layer's weighted integration to obtain the electricity demand forecast and its probability distribution, this approach is chosen. This includes considering the time correlation of the time series model layer with historical data; the larger the forecast time span, the weaker the timeliness of the historical data. A decay mechanism is designed for the time series model layer, including defining a time decay function with the following formula: α t =e -λ·t In the formula, t is the prediction time step, and λ is the decay coefficient, which is obtained by fitting historical errors; the defined time decay function is integrated into the set time series model basic weights; Considering the causal relationships in causal models that depend on external factors (weather, policy, economy), the explanatory power of the causal model increases as the volatility of these external factors increases. A volatility response mechanism is designed for the causal model layer, defining a volatility index with the following formula: In the formula, i represents the fluctuation factor, and the quantity is n, which comes from weather data, policy data, and economic data; The rolling standard deviation of the current factor i (e.g., fluctuations over a preset historical period) is given. Define the historical average volatility level of the current factor i (e.g., past daily volatility); integrate the defined volatility indicators into the set causal model base weights; Considering the use of uncertainty models to handle sudden risks (extreme weather, policy changes), their weight needs to be increased when high-risk scenarios occur. A risk triggering mechanism is designed for the uncertainty model layer, defining a multi-dimensional risk index with the following formula: γ=w1·R 外部因素异常度 +w2·R 模型预测不确定性 +w1·R 突发事件触发度 In the formula, w1, w2, and w3 are the weights of each indicator; R 外部因素异常度 The result is a weighted average of numerical anomalies standardized from weather, policy, and economic data. This includes pre-setting anomaly levels for different data ranges (e.g., temperatures exceeding 30°C have an anomaly level of 0.1, with a standardized range of [0,1]); electricity prices within the [X, Y] range have an anomaly level of 0.5, with a standardized range of [0,1]. Considering that models with similar parameters have relatively small differences in prediction uncertainty, R... 模型预测不确定性This is obtained by standardizing and weighting the historical prediction interval width and model divergence of the hybrid model under the condition that the hybrid model parameters are not optimized, based on historical adjacent time periods. Specifically, it is determined whether the historical prediction interval width of the hybrid model exceeds a preset multiple of the average historical prediction interval of the hybrid model under the condition that the hybrid model parameters are not adjusted for several historical adjacent time periods. If it exceeds this value, it is considered to have high uncertainty, and an uncertainty value of 1 is set accordingly; otherwise, it is set to 0. It is also determined by the degree of divergence of the historical predicted values ​​output by different model layers of the hybrid model under the condition that the hybrid model parameters are not optimized, and if the difference between the predicted values ​​output by any two model layers is greater than a preset value, the degree of divergence is considered to be severe, and an uncertainty value of 1 is set accordingly; otherwise, it is set to 0. 突发事件触发度 To quantify whether a historical emergency has been triggered based on real-time weather data, policy data, and economic data, the method involves statistically analyzing historical real-time weather data, policy data, and economic data, along with the corresponding numerical values ​​indicating the existence of such emergencies. This analysis determines whether the current real-time weather data, policy data, and economic data fall within the corresponding range of values ​​analyzed in the statistical analysis. If they do, the quantification value for whether a historical emergency has been triggered is 1; otherwise, the quantification value is 0. When γ is greater than a preset threshold, it indicates a high-risk scenario, and the calculated multidimensional risk index is integrated into the set causal model's basic weights. The dynamically adjusted weights corresponding to the time series model layer, causal model layer, and uncertain model layer are normalized. The normalized weights are then used for weighted ensemble to obtain the electricity demand forecast and its probability distribution. The specific normalization formula is as follows: Furthermore, the dynamically adjusted weights must satisfy the normalization condition, i.e.: w t +w c +w u =1, and: In the formula, w t w c w u These correspond to the dynamically adjusted weights for the time series model, causal model, and uncertainty model, respectively. These correspond to the basic weights of the time series model, causal model, and uncertainty model, respectively.

[0049] That is, for each future time period forecast, the dynamic weights of the forecast data are dynamically adjusted based on the real-time data collected in the current time period and the forecast data in the adjacent time periods. Then, the weighted integration of the adjusted weights is used to obtain the power demand forecast value and the probability distribution of the power demand forecast value.

[0050] In one specific embodiment, to further improve the robustness and accuracy of predictions, a real-time, periodic, or triggered model optimization strategy can be selected to optimize the prediction model under real-time, periodic, or triggered conditions; the method further includes: Obtain the actual value of electricity demand for future periods, and use the actual value of electricity demand and the corresponding electricity consumption data, weather data, policy data and economic data as the basis for incremental training data.

[0051] Considering the unique characteristics of hybrid models, which combine time-series, causal, and uncertainty models, parameter optimization may need to take into account the features of different model layers. For example, time-series models may require more frequent updates, while causal models may be affected by policy and economic factors and have a lower update frequency. Uncertainty models may require updates triggered under high-risk scenarios. Therefore, a hierarchical dynamic optimization strategy is designed to perform incremental training for different types of model layers according to the matching hierarchical dynamic optimization strategy in order to complete the model parameter optimization.

[0052] Among them, the hierarchical dynamic optimization strategy that matches the time series model layer is to combine real-time incremental training with periodic incremental training. Due to the sensitivity of the LSTM model to time series data, new actual power demand data is collected in real time and incorporated into the model for incremental training. Furthermore, at regular intervals (such as once a week or once a month), all historical data can be used for full training to ensure that the model can fully learn the long-term changing patterns of power demand.

[0053] Among them, the hierarchical dynamic optimization strategy that matches the causal model layer adopts a combination of triggered incremental training and periodic incremental training. When the difference between the current weather, policy or economic data and the previous time period is greater than the corresponding preset difference value, the incremental training of the XGBoost model is triggered; it can also periodically use all historical data for a full training.

[0054] The hierarchical dynamic optimization strategy that matches the uncertain model layer is to combine periodic incremental training with triggered training. When the deviation between the actual power demand and the model prediction is detected to exceed the preset deviation, incremental training of the uncertain model is triggered. Alternatively, a full training can be performed periodically using all historical data.

[0055] In a specific embodiment, to better adapt to different market environments and policy requirements, and to achieve a flexible balance between energy allocation strategies and energy conservation, emission reduction, and cost control, thereby meeting the dual objectives of low-carbon environmental protection and maximizing economic benefits, the method further includes: The current guidance scenario type is determined based on the obtained policy and economic data, including: if the current carbon trading price is greater than the preset carbon trading price (80 yuan / ton), it is identified as a strong emission reduction guidance scenario; if the current electricity price is less than the preset electricity price (90% of the benchmark electricity price), it is identified as a cost-priority guidance scenario; if both the current carbon trading price and the current electricity price are greater than or less than the preset electricity price, it is identified as a strong emission reduction guidance scenario; if neither the current carbon trading price nor the current electricity price is greater than or less than the preset electricity price, it is identified as a balanced optimization guidance scenario. Based on the current guidance scenario type, the weight settings of total carbon emissions and electricity costs in the objective function are dynamically adjusted. Specifically, for the strong emission reduction guidance scenario type, the weight range of total carbon emissions (0.7-0.9) should be set to be greater than the weight range of electricity costs (0.1-0.3); for the cost priority guidance scenario type, the weight range of total carbon emissions (0.3-0.5) should be set to be less than the weight range of electricity costs (0.5-0.7); for the balanced optimization guidance scenario type, the maximum weight difference between the weight range of total carbon emissions (0.5-0.6) and the weight range of electricity costs (0.4-0.5) should be less than the preset weight difference.

[0056] One specific embodiment involves acquiring electricity consumption, weather, policy, and economic data, using a hybrid forecasting model to predict electricity demand, and then considering carbon emission quotas. A modified objective function is used to replace the original objective function to construct a multi-objective constrained optimization model to solve for energy-carbon linkage optimization allocation strategies, achieving high-precision prediction of electricity demand, reducing supply-demand imbalances, and lowering energy waste. The method also includes: Considering the setting of carbon emission quotas, the carbon emission quotas for users in future periods are obtained, and a modified objective function is selected to replace the original objective function. A multi-objective constrained optimization model is then constructed to solve for the energy-carbon linkage allocation strategy in future periods. The modified objective function aims to minimize the carbon quota penalty and the normalized electricity cost, and the formula is: minα·Cost total '+β·Carbon penalty In the formula, Cost total ’ Cost represents the normalized cost of electricity. total Cost of electricity min Cost is the lowest historical cost for users. max The user's highest historical cost; Carbon penaltyCarbon as a penalty for carbon quotas cap γ represents the user's carbon emission allowance for the future period; γ is the penalty coefficient, such as 1000 yuan / kgCO2, simulating the carbon trading penalty; α and β are the corresponding weights.

[0057] like Figure 2 As shown in the figure, this application discloses a power line carrier energy-carbon linkage control system based on hybrid prediction, specifically including: The power data acquisition module 101 is used to acquire power consumption data, weather data, policy data, and economic data transmitted via power line carrier communication. The electricity demand forecasting module 102 is used to input the acquired electricity consumption data, weather data, policy data, and economic data into a hybrid forecasting model to predict the electricity demand forecast for future periods. The hybrid forecasting model adopts a hierarchical forecasting framework, including a basic decision layer and a decision fusion layer. The basic forecasting layer is used in advance to obtain the initial electricity demand forecast and its probability distribution. The basic forecasting layer is equipped with a time series model layer based on LSTM, a causal model layer based on XGBoost, and an uncertainty model layer based on a Bayesian model. The decision fusion layer is then used to weightedly integrate and obtain the electricity demand forecast and its probability distribution. The carbon allocation strategy acquisition module 103 is used to construct a multi-objective constrained optimization model to solve for and obtain the carbon-linked allocation strategy for future periods, including the amount of electricity purchased, the amount of clean energy supplied, the amount of energy storage charged and discharged, and the state of charge of energy storage in future periods. The multi-objective constrained optimization model takes minimizing total carbon emissions and electricity costs as its objective function, and sets the following constraints: the amount of electricity purchased, the amount of clean energy supplied, the amount of energy storage charged and discharged, and the predicted value of electricity demand in future periods must meet the supply and demand balance; the amount of clean energy supplied in future periods must not exceed the upper limit of clean energy supply; the amount of electricity purchased in future periods must not exceed the upper limit of grid electricity purchase; and the energy storage capacity in future periods must not exceed the preset energy storage capacity. The energy and carbon allocation strategy execution module 104 is used to generate electricity purchase instructions, electricity generation instructions and electricity storage and release instructions according to the energy and carbon linkage allocation strategy for future periods, and send them to the electricity purchase client, electricity production capacity device and electricity storage and release device through power carrier communication to control electricity purchase, production capacity and storage and release.

[0058] In a specific embodiment, the power demand forecasting module 102 in the system is further used to construct and train a dynamic adjustment model for forecast periods. The dynamic adjustment model for forecast periods employs a deep learning algorithm. The input to the model is real-time acquired power consumption area information and seasonal information transmitted via power line carrier communication, as well as key power consumption data and key weather data extracted based on the power consumption area information and seasonal information. The output is the length of the forecast period. The model is trained using historically acquired power consumption area information and seasonal information, key power consumption data and key weather data extracted based on historically acquired power consumption area information and seasonal information, and the preset time length range within which the predicted power demand value for the future period is less than the actual power demand value. The key power consumption data and key weather data extracted based on the acquired power consumption area information and seasonal information are extracted according to rules that preset key power consumption characteristics for different power consumption areas and preset key weather information for different area information and seasonal information. The real-time acquired electricity consumption area information and seasonal information, as well as key electricity consumption data and key weather data extracted based on the real-time acquired electricity consumption area information and seasonal information, are input into the dynamic adjustment model for the forecast period to obtain the forecast period length; the future period length in the forecast value of electricity demand for the future period is adjusted according to the forecast period length.

[0059] In one specific embodiment, the power demand forecasting module 102 in the system is further configured to select the use of a set decision fusion layer to dynamically weighted integrate and obtain the power demand forecast value and the probability distribution of the power demand forecast value instead of using the set decision fusion layer to weighted integrate and obtain the power demand forecast value and the probability distribution of the power demand forecast value.

[0060] In one specific embodiment, the system further includes: a power demand forecasting and optimization module 105, used to obtain the actual power demand value for future periods, and use the actual power demand value and the corresponding electricity consumption data, weather data, policy data, and economic data as the basis for incremental training data; designing a hierarchical dynamic optimization strategy, and performing incremental training on different types of model layers according to the matching hierarchical dynamic optimization strategy to complete the optimization of model parameters; wherein, the hierarchical dynamic optimization strategy matching the time series model layer adopts a combination of real-time incremental training and periodic incremental training, the hierarchical dynamic optimization strategy matching the causal model layer adopts a combination of triggered incremental training and periodic incremental training, and the hierarchical dynamic optimization strategy matching the uncertain model layer adopts a combination of periodic incremental training and triggered training.

[0061] In one specific embodiment, the carbon allocation strategy acquisition module 103 in the system is further configured to determine the current guidance scenario type based on the acquired policy data and economic data, including: if the current carbon trading price is greater than the preset carbon trading price, it is identified as a strong emission reduction guidance scenario type; if the current electricity price is less than the preset electricity price, it is identified as a cost-priority guidance scenario type; if both the current carbon trading price and the current electricity price are greater than the preset carbon trading price are met, it is identified as a strong emission reduction guidance scenario type; if neither the current carbon trading price nor the current electricity price is greater than the preset carbon trading price is met, it is identified as a balanced optimization guidance scenario type; based on the determined current guidance scenario type, the weight settings of total carbon emissions and electricity costs in the objective function are dynamically adjusted; wherein, for the strong emission reduction guidance scenario type, the weight range of total carbon emissions should be set to be greater than the weight range of electricity costs; for the cost-priority guidance scenario type, the weight range of total carbon emissions should be set to be less than the weight range of electricity costs; for the balanced optimization guidance scenario type, the maximum weight difference between the weight range of total carbon emissions and the weight range of electricity costs should be less than the preset weight difference.

[0062] In one specific embodiment, the energy and carbon allocation strategy acquisition module 103 in the system is further used to consider the carbon emission quotas set, acquire the user's carbon emission quotas for future periods, select a modified objective function to replace the original objective function, and then construct a multi-objective constrained optimization model to solve and obtain the energy and carbon linkage allocation strategy for future periods.

[0063] The embodiment of the present application also discloses a computer-readable storage medium.

[0064] Specifically, the computer-readable storage medium stores a computer program that can be loaded by a processor and executed, such as the above-described hybrid prediction-based power line carrier energy carbon linkage control method. The computer-readable storage medium includes, for example, various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0065] The embodiment of the present application also discloses a computer device.

[0066] Specifically, the computer device includes a memory and a processor, and the memory stores a computer program that can be loaded by the processor and executed by the aforementioned power carrier energy carbon linkage control method based on hybrid prediction.

[0067] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.

Claims

1. A power line carrier energy-carbon linkage control method based on hybrid prediction, characterized in that, include: To acquire electricity consumption data, weather data, policy data, and economic data transmitted via power line carrier communication; The acquired electricity consumption data, weather data, policy data, and economic data are input into a hybrid forecasting model to predict electricity demand for future periods. The hybrid forecasting model adopts a hierarchical forecasting framework, including: a basic forecasting layer and a decision fusion layer; the basic forecasting layer is used in advance to obtain the initial electricity demand forecast and its probability distribution; the basic forecasting layer is equipped with a time series model layer based on LSTM, a causal model layer based on XGBoost, and an uncertainty model layer based on a Bayesian model; and the decision fusion layer is used to obtain the electricity demand forecast and its probability distribution through weighted integration. A multi-objective constrained optimization model is constructed to solve for the future energy-carbon linkage allocation strategy, including the future electricity purchase amount, clean energy supply amount, energy storage charging and discharging amount, and energy storage state of charge. The multi-objective constrained optimization model takes minimizing total carbon emissions and electricity costs as the objective function, and sets the following constraints: the future electricity purchase amount, clean energy supply amount, energy storage charging and discharging amount and the predicted value of electricity demand in the future period meet the supply and demand balance; the future clean energy supply amount does not exceed the upper limit of clean energy supply; the future electricity purchase amount does not exceed the upper limit of grid electricity purchase; and the future energy storage capacity does not exceed the preset energy storage capacity. According to the future energy-carbon linkage allocation strategy, corresponding power purchase instructions, power generation instructions, and power storage and release instructions are generated and sent to the power purchase client, power production device, and power storage and release device via power carrier communication to control power purchase, production, and storage and release.

2. The power line carrier energy-carbon linkage control method based on hybrid prediction according to claim 1, characterized in that, Also includes: A dynamic adjustment model for predictive time periods is constructed and trained. This model employs a deep learning algorithm. The input to the model consists of real-time acquired electricity consumption area information and seasonal information transmitted via power line carrier communication, as well as key electricity consumption data and key weather data extracted based on these information. The output is the length of the predicted time period. The model is trained using historically acquired electricity consumption area information and seasonal information, key electricity consumption data and key weather data extracted based on these information, and a preset time length range within which the predicted and actual electricity demand values ​​for future periods have an error less than a preset error value. The key electricity consumption data and key weather data extracted based on the acquired electricity consumption area information and seasonal information are extracted according to rules that preset key electricity consumption characteristics for different electricity consumption areas and preset key weather information for different area and season information. The real-time acquired electricity consumption area information and seasonal information, as well as key electricity consumption data and key weather data extracted based on the real-time acquired electricity consumption area information and seasonal information, are input into the dynamic adjustment model for the forecast period to obtain the forecast period length; the future period length in the forecast value of electricity demand for the future period is adjusted according to the forecast period length.

3. The power line carrier energy-carbon linkage control method based on hybrid prediction according to claim 1, characterized in that, Also includes: Instead of using the set decision fusion layer to obtain the power demand forecast and its probability distribution through dynamic weighted integration, we choose to use the set decision fusion layer to obtain the power demand forecast and its probability distribution. include: A decay mechanism is designed for the time-series model layer, including: defining a time decay function, with the formula: α t ＝e -λ·t In the formula, t is the prediction time step, and λ is the decay coefficient, which is obtained by fitting historical errors; the defined time decay function is integrated into the set time series model basic weights; A volatility response mechanism is designed for the causal model layer, and a volatility index is defined with the following formula: In the formula, i represents the fluctuation factor, and the quantity is n, which comes from weather data, policy data, and economic data; Let i be the rolling standard deviation of the current factor i. Define the historical average volatility level of current factor i; integrate the defined volatility index into the set causal model base weights; For the uncertain model layer, a risk triggering mechanism is designed, and a multidimensional risk index is defined with the following formula: γ=w1·R 外部因素异常度 +w2·R 模型预测不确定性 +w1·R 突发事件触发度 In the formula, w1, w2, and w3 are the weights of each indicator; R 外部因素异常度 R is obtained by weighting the values ​​after standardizing numerical anomalies based on weather data, policy data, and economic data; 模型预测不确定性 R is obtained by standardizing and weighting the prediction interval width and model divergence of the mixture model based on historical adjacent time periods and without optimization of the mixture model parameters; 突发事件触发度 This is a quantitative value used to determine whether historical emergencies are triggered based on real-time weather data, policy data, and economic data; when γ is greater than a preset threshold, the calculated multidimensional risk index is integrated into the set causal model base weights. The dynamically adjusted weights of the time series model layer, causal model layer, and uncertain model layer are normalized, and the corresponding weights are weighted and integrated to obtain the electricity demand forecast and the probability distribution of the electricity demand forecast.

4. The power line carrier energy-carbon linkage control method based on hybrid prediction according to claim 1, characterized in that, Also includes: Obtain the actual value of electricity demand for future periods, and use the actual value of electricity demand and the corresponding electricity consumption data, weather data, policy data and economic data as the basis for incremental training data; A hierarchical dynamic optimization strategy is designed to perform incremental training for different types of model layers according to the matching hierarchical dynamic optimization strategy to complete the model parameter optimization. Specifically, the hierarchical dynamic optimization strategy matching the temporal model layer adopts a combination of real-time incremental training and periodic incremental training, the hierarchical dynamic optimization strategy matching the causal model layer adopts a combination of triggered incremental training and periodic incremental training, and the hierarchical dynamic optimization strategy matching the uncertain model layer adopts a combination of periodic incremental training and triggered training.

5. The power line carrier energy-carbon linkage control method based on hybrid prediction according to claim 1, characterized in that, Also includes: The current guidance scenario type is determined based on the obtained policy and economic data, including: if the current carbon trading price is greater than the preset carbon trading price, it is identified as a strong emission reduction guidance scenario; if the current electricity price is less than the preset electricity price, it is identified as a cost-priority guidance scenario; if both the current carbon trading price and the current electricity price are greater than or less than the preset electricity price, it is identified as a strong emission reduction guidance scenario; if neither the current carbon trading price nor the current electricity price is greater than or less than the preset electricity price, it is identified as a balanced optimization guidance scenario. Based on the current guidance scenario type, the weight settings of total carbon emissions and electricity costs in the objective function are dynamically adjusted. Specifically, for the strong emission reduction guidance scenario type, the weight range of total carbon emissions should be set to be greater than the weight range of electricity costs; for the cost priority guidance scenario type, the weight range of total carbon emissions should be set to be less than the weight range of electricity costs; and for the balanced optimization guidance scenario type, the maximum weight difference between the weight range of total carbon emissions and the weight range of electricity costs should be less than the preset weight difference.

6. The power line carrier energy-carbon linkage control method based on hybrid prediction according to claim 1, characterized in that, Also includes: Considering the setting of carbon emission quotas, obtain the carbon emission quotas of users in the future period, select the modified objective function to replace the original objective function, and then construct a multi-objective constrained optimization model to solve and obtain the energy-carbon linkage allocation strategy in the future period. The modified objective function aims to minimize the carbon quota penalty and the normalized electricity cost, and the formula is as follows: minα·Cost total ′+β·Carbon penalty In the formula, Cost total ’ Cost represents the normalized cost of electricity. total Cost of electricity min Cost is the lowest historical cost for users. max The user's highest historical cost; Carbon penalty Carbon as a penalty for carbon quotas cap Carbon emission allowances for users in the future; γ is the penalty coefficient; α and β are the weights of the corresponding terms.

7. The power line carrier energy-carbon linkage control method based on hybrid prediction according to claim 1, characterized in that, Also includes: Before the generated power purchase instruction, power generation instruction, and generated power storage instruction are sent to the power purchase client, power production unit, and power storage unit respectively via power line carrier communication, each instruction is pre-encoded. Upon receiving the instruction verification results from the power purchase client, power production unit, and power storage unit, and the result of instruction verification failure, retransmission is selected based on the result of instruction verification failure.

8. A power line carrier energy-carbon linkage control system based on hybrid prediction, characterized in that, include: The power data acquisition module is used to acquire electricity consumption data, weather data, policy data, and economic data transmitted via power line carrier communication. The electricity demand forecasting module is used to input the acquired electricity consumption data, weather data, policy data, and economic data into a hybrid forecasting model to predict the electricity demand forecast for future periods. The hybrid forecasting model adopts a hierarchical forecasting framework, including: a basic decision layer and a decision fusion layer; the basic forecasting layer is used in advance to obtain the initial electricity demand forecast and the probability distribution of the initial electricity demand forecast; the basic forecasting layer is set with a time series model layer based on LSTM, a causal model layer based on XGBoost and an uncertain model layer based on Bayesian model, and then the set decision fusion layer is used to weighted integrate to obtain the electricity demand forecast and the probability distribution of the electricity demand forecast; The energy and carbon allocation strategy acquisition module is used to construct a multi-objective constrained optimization model to solve for and obtain the energy and carbon linkage allocation strategy for future periods, including the amount of electricity purchased, the amount of clean energy supplied, the amount of energy storage charged and discharged, and the state of charge of energy storage for future periods. The multi-objective constrained optimization model takes minimizing total carbon emissions and electricity costs as its objective function, and sets the following constraints: the amount of electricity purchased, the amount of clean energy supplied, the amount of energy storage charged and discharged for future periods must be in supply and demand balance with the predicted value of electricity demand for future periods; the amount of clean energy supplied for future periods must not exceed the upper limit of clean energy supply; the amount of electricity purchased for future periods must not exceed the upper limit of grid electricity purchase; and the energy storage capacity for future periods must not exceed the preset energy storage capacity. The energy and carbon allocation strategy execution module is used to generate electricity purchase instructions, electricity generation instructions, and electricity storage and release instructions according to the energy and carbon linkage allocation strategy for future periods. These instructions are then sent to the electricity purchase client, the electricity production capacity device, and the electricity storage and release device via power line carrier communication to control electricity purchase, production capacity, and storage and release.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the method as described in any one of claims 1 to 7.

10. A computer device, characterized in that, The computer device includes a memory, a processor, and a program stored in and executable on the memory, the program being executed by the processor to implement the steps of the method as described in any one of claims 1 to 7.

Images