Customer carbon footprint prediction method based on electric energy supply and demand prediction and electric power transaction contract

Abstract

The invention is suitable for the technical field of carbon emission, and provides a customer carbon footprint prediction method based on electric energy supply and demand prediction and a power transaction contract, and the method comprises the following steps: collecting historical data, training an LSTM-AdaBoost neural network model based on the historical data, and carrying out the prediction of the output of a new energy station and the load of a power distribution and utilization user. Future weather forecast parameters are collected, and the LSTM-AdaBoost neural network model is used to predict the output of the new energy station and the load of the power distribution and utilization user in the future; based on the power transaction contract and the transaction curve, power generation side output and power distribution and utilization client load matching is carried out; and based on the thermal power generating unit carbon emission factor data, carrying out future carbon footprint change trend prediction on the power distribution and utilization clients. According to the method, the change trend prediction of the power consumption carbon footprint of the power distribution and utilization clients in the future one week is realized, enterprises can be guided to preferentially consume new energy for power generation, and power carbon demand response is developed through scientific production arrangement and power utilization behavior optimization.

Description

Technical Field

[0001] The present invention relates to the field of carbon emission technology, and in particular to a method for predicting customer carbon footprint based on power supply and demand forecasting and power trading contracts. Background Art

[0002] Electricity carbon footprint accounting quantifies indirect carbon emissions in the product production process through the grid carbon emission factor, that is, the carbon content per unit of electricity. The spatiotemporal distribution characteristics of the grid carbon emission factor directly affect the accuracy of the accounting results. Achieving short-term carbon footprint prediction for electricity consumption is of great significance for guiding energy transformation, promoting clean energy consumption, and corporate low-carbon decision-making. However, the traditional annual statistics of electricity carbon factor data cannot meet the current requirements for real-time, accuracy, and refinement of carbon footprint accounting.

[0003] The calculation of carbon footprint based on the national average carbon emission factor of the power grid is to calculate indirect carbon emissions by multiplying the load and the carbon emission factor of the power grid. At present, the annual power generation, power generation fuel consumption, average low calorific value of power generation fuel, inter-regional power transmission and other key data are comprehensively considered. According to the distribution status of the power grid, the average carbon emission factor is provided from three levels: national power grid, regional power grid and provincial power grid for calculating the carbon footprint of user-end power consumption.

[0004] However, the scope of application of the above scheme is limited: the average carbon emission factor of the power grid is more suitable for the macro analysis and evaluation of the overall carbon emissions of the power system, but its applicability may be limited when it comes to the carbon emission analysis of a certain area, a certain type of equipment or a certain time period. Difficult to cope with complex situations: The power system is a complex system that contains multiple types of equipment and multiple forms of energy. When faced with complex situations (such as renewable energy access, virtual power plant peak load, etc.), the carbon footprint accounting method of the average carbon emission factor of the power grid may be difficult to provide accurate carbon emission analysis and evaluation. Insufficient dynamics: The operating state of the power system is dynamically changing, and the average carbon emission factor of the power grid is often based on a certain historical period when it is announced by the relevant departments. This static analysis may not accurately reflect the impact of dynamic changes in the development and operation of the power system on carbon emissions, nor can it meet the power consumption analysis needs of distribution and power customers to reduce carbon footprints.

[0005] Therefore, it is necessary to provide a customer carbon footprint prediction method based on electricity supply and demand forecasting and electricity trading contracts to solve the above problems. Summary of the invention

[0006] In view of the deficiencies in the prior art, the purpose of the present invention is to provide a customer carbon footprint prediction method based on electricity supply and demand forecasting and electricity trading contracts to solve the problems existing in the above-mentioned background technology.

[0007] The present invention is implemented as follows: a method for predicting customer carbon footprint based on power supply and demand forecasting and power trading contracts, the method comprising the following steps:

[0008] Collect historical data, including the medium- and long-term power trading contracts between new energy stations and electricity users, agreed trading curves, station weather conditions and operating conditions, and train the LSTM-AdaBoost neural network model based on the historical data to predict the output of new energy stations and the load of power distribution users;

[0009] Collect future weather forecast parameters and use the LSTM-AdaBoost neural network model to predict the future output of new energy stations and the load size of power distribution users;

[0010] Based on the power trading contract and trading curve, the power generation output and the load of the power distribution and consumption customers are matched;

[0011] Based on the carbon emission factor data of thermal power units, the future carbon footprint change trend of power distribution customers is predicted.

[0012] As a further solution of the present invention: the step of training the LSTM-AdaBoost neural network model based on historical data specifically includes:

[0013] Based on the AdaBoost ensemble algorithm, multiple LSTM networks are trained serially as base learners;

[0014] Clean and normalize the historical data to obtain the training data set and the test data set;

[0015] The LSTM-AdaBoost neural network model is trained. During the model training process, a three-layer LSTM network model is used, each layer contains 128 units, and the mean square error is used as the loss function. The Adam optimizer is used for optimization. In order to prevent the model from overfitting, the early stopping method and Dropout technology are used during the training process.

[0016] As a further solution of the present invention: when training multiple LSTM networks as base learners, the training process starts with initializing the weights of the training set and setting the number of base learners T, and then the training sets are input into the LSTM network one by one for training. After each round of training, the relative error eti of the sample, the regression error rate εt of the LSTM network and the weight coefficient αt of the base learner are calculated, the weight distribution of the training set is updated, and the above steps are repeated until the preset number of base learners T is reached, and all the trained base learners are linearly combined by their weight coefficients to form an LSTM-AdaBoost neural network model.

[0017] As a further solution of the present invention: the step of training the LSTM-AdaBoost neural network model specifically includes:

[0018] Initialize the weight vector, and the weight distribution of the training data is initialized as:

[0019] For k = 1, 2, ..., K: When the weight distribution is D k When , the training data is trained through the LSTM neural network to obtain the predictor based on the LSTM model: h k :S→Y;

[0020] Calculate h k Prediction error on the training dataset: Where M = sup i (|h k (x i )-y i |), the output interval of the function is [0,1]; calculate the total error of the training sample set: Calculate the current predictor h k The coefficient of in Update the weight distribution of the training dataset: in

[0021] The loop ends and the connection weight W of the LSTM predictor is recorded. 1 ,w 2 ,…,w K ),in

[0022] Construct the final predictor and integrate the trained predictors according to the connection weights to obtain the final strong predictor: h(x) = w 1 h 1 (x)+w 2 h 2 (x)+…+w K h K (x).

[0023] As a further solution of the present invention: the step of matching the power generation side output with the power distribution customer load specifically includes:

[0024] Based on the contract matrix, a constraint matrix form of supply and demand matching is constructed. The matrix P represents all power transaction relationships of supply and demand matching, where the rows represent power demand parties and the columns represent power suppliers.

[0025] Based on the total demand, total supply and sources of power shortage, constraints are determined to match the output of the power generation side with the load of the power distribution customers.

[0026] As a further solution of the present invention: the step of predicting the future carbon footprint change trend of power distribution customers based on the carbon emission factor data of thermal power units specifically includes:

[0027] Based on the carbon emission factor data of thermal power units, combined with the real-time matching of power generation output and demand side load, the user's carbon footprint is calculated:

[0028] The carbon emissions of various types of electricity are summarized to predict the carbon footprint of power distribution and power consumption customers. The calculation formula is: Total carbon footprint = ∑ electricity i × EF i , where i represents different types of electricity, the amount of electricity i is the amount of electricity consumed by the corresponding type, and EF i is the corresponding emission factor.

[0029] As a further solution of the present invention: the electricity types include photovoltaic power, wind power and thermal power. For photovoltaic power and wind power, the emission factor is 0; for thermal power, EF is the real-time emission factor of the thermal power unit.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] The present invention adopts a carbon footprint prediction method for distribution and user customers based on electricity supply and demand forecasts and electricity trading contracts to realize the prediction of changing trends in the carbon footprint of electricity consumption of distribution and user customers in the next week. This is conducive to guiding enterprises to give priority to the consumption of new energy power generation, respond to electricity-carbon demand through scientific production arrangements and optimization of electricity consumption behaviors, and reduce comprehensive electricity costs and the negative environmental impact of carbon emissions. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 The present invention is a flow chart of a customer carbon footprint prediction method based on power supply and demand forecasting and power trading contracts.

[0033] Figure 2 This is a schematic diagram of LSTM-AdaBoost training results;

[0034] Figure 3 This is a schematic diagram of LSTM-AdaBoost prediction results;

[0035] Figure 4 This is a graph of real-time supply-load forecast matching results for summer power contracts;

[0036] Figure 5 This is a graph of real-time supply-load forecast matching results for winter power contracts;

[0037] Figure 6 Carbon footprint prediction results for electricity distribution customers in summer;

[0038] Figure 7 This is a graph showing the carbon footprint prediction results for electricity distribution customers in winter. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0040] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments.

[0041] like Figure 1 As shown, an embodiment of the present invention provides a method for predicting customer carbon footprint based on power supply and demand forecasting and power trading contracts, the method comprising the following steps:

[0042] S100, collecting historical data, wherein the historical data includes the medium- and long-term power trading contracts between the new energy station and the power users and the agreed trading curves, the weather conditions of the station and the operating conditions, and training the LSTM-AdaBoost neural network model based on the historical data to predict the output of the new energy station and the load of the power distribution users;

[0043] S200 collects weather forecast parameters for the next week and uses the LSTM-AdaBoost neural network model to predict the output of new energy stations and the load size of power distribution users in the next week;

[0044] S300, based on power trading contracts and trading curves, matches the output of power generation side with the load of power distribution and consumption customers;

[0045] S400, based on the carbon emission factor data of thermal power units, predicts the changing trend of the carbon footprint of power distribution customers in the next week.

[0046] In an embodiment of the present invention, the step of training the LSTM-AdaBoost neural network model based on historical data specifically includes: serially training multiple LSTM networks as base learners based on the AdaBoost integrated algorithm; cleaning and normalizing the historical data to obtain high-quality training data sets and test data sets. In the model training process, a three-layer LSTM network model is used, each layer contains 128 units, and the mean square error is used as the loss function, and the Adam optimizer is used for optimization. In order to prevent the model from overfitting, the early stopping method and Dropout technology are used in the training process.

[0047] In an embodiment of the present invention, when multiple LSTM networks are trained as base learners, the training process starts with initializing the weights of the training set and setting the number of base learners T, and then the training set is input into the LSTM network one by one for training. After each round of training, the relative error eti of the sample, the regression error rate εt of the LSTM network, and the weight coefficient αt of the base learner are calculated, and the weight distribution of the training set is updated so that the samples predicted incorrectly in the next round of training obtain higher weights, thereby guiding the subsequent LSTM network to pay more attention to these samples, and repeating this step until the preset number of base learners T is reached, and all the trained base learners are linearly combined through their weight coefficients to form an LSTM-AdaBoost neural network model.

[0048] In the embodiment of the present invention, the step of training the LSTM-AdaBoost neural network model specifically includes:

[0049] Initialize the weight vector, and the weight distribution of the training data is initialized as:

[0050] For k = 1, 2, ..., K: When the weight distribution is D k When , the training data is trained through the LSTM neural network to obtain the predictor based on the LSTM model: h k :S→Y;

[0051] Calculate h k Prediction error on the training dataset: Where M = sup i (|h k (x i )-y i |), the output interval of the function is [0,1]; calculate the total error of the training sample set: Calculate the current predictor h k The coefficient of in Update the weight distribution of the training dataset: in

[0052] The loop ends and the connection weight W of the LSTM predictor is recorded. 1 ,w 2 ,…,w K ),in

[0053] Construct the final predictor and integrate the trained predictors according to the connection weights to obtain the final strong predictor: h(x) = w 1 h 1 (x)+w 2 h 2(x)+…+w K h K (x).

[0054] After completing the training and optimization of the LSTM-AdaBoost integrated model, the LSTM-AdaBoost integrated model is used to predict the output of new energy stations and the load size of distribution and power users in the next week, using the clear medium- and long-term electricity trading contracts and trading curve data between market entities such as power generation companies and power users, power sales companies and emerging entities obtained from the power market, as well as the meteorological parameters for the next week given by the weather forecast service.

[0055] It should be noted that in the power system, power distribution customers include different types such as residential, commercial, industrial and special-purpose users. Each type of user has significant differences in electricity demand, power load characteristics and seasonal changes. Residential users' electricity consumption has obvious time and seasonal characteristics, commercial users are affected by the time and nature of business activities, and industrial users' electricity consumption is usually large and stable. The electricity consumption characteristics of special-purpose users such as public facilities and agricultural users are affected by working hours and production cycles. By decomposing the power trading contract, the load characteristics of the power distribution customers are analyzed in detail, the power supply side and the load side demand are connected, and the supply and demand relationship is accurately matched. The steps for matching the output of the power generation side with the load of the power distribution customers specifically include:

[0056] Based on the contract matrix, a constraint matrix of supply and demand matching is constructed. The matrix P represents all power transaction relationships of supply and demand matching, where the rows represent power demand parties and the columns represent power suppliers. Assume that P ij Represents the power (MW) purchased by the i-th enterprise (or power demander) from the j-th power plant (or power supplier). The matrix P represents all power trading relationships matching supply and demand, where: rows represent power demanders (such as industrial enterprises, commercial electricity, enterprise parks, power sales agents, etc.), and columns represent power suppliers (such as wind power plants, photovoltaic plants, thermal power plants, etc.). Assuming there are m power demanders and n power suppliers, the matrix P is represented as:

[0057]

[0058] Based on the total demand, total supply and the source of power shortage, the constraints are determined to match the output of the power generation side with the load of the power distribution customers. For each power demander i, its total demand should be equal to the sum of the power purchased from each power plant: D i represents the total demand of power demander i. For each power supplier j, its total supply should be equal to the sum of the power it allocates to each demander, and is limited by its maximum supply capacity C j : Cj Represents the supply capacity of power supplier j. Shortage power source constraint: When the output of new energy is insufficient, thermal power is used as the source of shortfall power. A constraint condition is set, and the output of thermal power should meet the shortfall:

[0059]

[0060] Supply and demand matching: Supply and demand matching can be expressed as a linear programming problem. The goal is to minimize the gap between supply and demand, satisfying all the above constraints:

[0061] minimize

[0062] Constraints:

[0063] In the embodiment of the present invention, the step of predicting the future carbon footprint change trend of the power distribution and user customers based on the carbon emission factor data of the thermal power units specifically includes: calculating the user's carbon footprint based on the carbon emission factor data of the thermal power units, combined with the real-time matching of the power generation side output and the demand side load: summarizing the carbon emissions of various types of electricity, predicting the carbon footprint of the power distribution and user customers, and the calculation formula is: total carbon footprint = ∑ electricity i × EF i , where i represents different types of electricity, the amount of electricity i is the amount of electricity consumed by the corresponding type, and EF i The electricity types include photovoltaic power, wind power and thermal power. For photovoltaic power and wind power, the emission factor is 0; for thermal power, EF is the real-time emission factor of the thermal power unit.

[0064] See also Figures 2 to 7 , the following is an actual example analysis:

[0065] Matching the output of new energy stations with the load of power distribution and consumption customers: In the model training stage, we selected 1200-day medium- and long-term power trading contracts and agreed trading curves, weather conditions (temperature, humidity, solar radiation intensity, wind speed, precipitation), new energy station operating conditions, power grid load and other time series samples as training sets to complete the LSTM-AdaBoost output and load forecasting model training. The root mean square error tended to 0.01 stably in 800 iterations, and the RMSE of the prediction results was about 0.072, which shows that the model results fit well.

[0066] Output-load matching results based on power trading contracts: Assume that there are power trading contract behaviors shown in the following table in this example:

[0067]

[0068]

[0069] See also Figure 4 and Figure 5 , and gives the output-load matching results of each power plant and electricity user, realizing the supply and demand relationship prediction based on the power trading contract. In the real-time supply-load matching results of the summer and winter green power contracts, the data shows that there are obvious seasonal fluctuations in the output of new energy.

[0070] Carbon footprint prediction results of power distribution customers: According to the above method, the total amount of electricity of new energy power in the power load is quantified based on the power supply and demand forecast, and the specific proportion of thermal power generation in the terminal power consumption is obtained. Subsequently, the carbon footprint of power consumption of power distribution customers is predicted using the real-time emission factor of thermal power units. Finally, the daily carbon footprints of different industrial enterprises, commercial customers, and power sales companies during one week in summer and winter are shown in the figure. In this example, industrial enterprise 1 and wind farm station 1, industrial enterprise 3 and photovoltaic farm station 2 signed green power transactions respectively, and the power load is mainly based on new energy, so the power carbon footprint of these two types of users is considered to be 0. Since the power generation efficiency of photovoltaic technology in winter is lower than that in summer, and winter coincides with the peak period of commercial and residential electricity demand, the carbon emissions of commercial and residential electricity in winter increase significantly, reflecting the interaction between seasonal power consumption patterns and fluctuations in renewable energy supply.

[0071] The embodiment of the present invention adopts a carbon footprint prediction method for distribution and user customers based on electricity supply and demand forecasts and electricity trading contracts to achieve prediction of changing trends in the carbon footprint of electricity consumption of distribution and user customers in the coming week. This is conducive to guiding enterprises to give priority to the consumption of new energy power generation, respond to electricity and carbon demand through scientific production arrangements and optimization of electricity consumption behaviors, reduce comprehensive electricity costs and the negative environmental impact of carbon emissions, and enhance the green value and international competitiveness of products.

[0072] The above only describes in detail the preferred embodiments of the present invention, which is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

[0073] It should be understood that, although each step in the flow chart of each embodiment of the present invention is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0074] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0075] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the disclosure in the specification and examples. This application is intended to cover any variations, uses or adaptations of the present disclosure, which follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the claims.

Claims

1. A customer carbon footprint prediction method based on power supply and demand forecasting and power trading contracts, characterized in that: The method comprises the following steps: Collect historical data, including the medium- and long-term power trading contracts between new energy stations and electricity users, agreed trading curves, station weather conditions and operating conditions, and train the LSTM-AdaBoost neural network model based on the historical data to predict the output of new energy stations and the load of power distribution users; Collect future weather forecast parameters and use the LSTM-AdaBoost neural network model to predict the future output of new energy stations and the load size of power distribution users; Based on the power trading contract and trading curve, the power generation output and the load of the power distribution and consumption customers are matched; Based on the carbon emission factor data of thermal power units, the future carbon footprint change trend of power distribution customers is predicted.

2. The customer carbon footprint prediction method based on power supply and demand forecasting and power trading contracts according to claim 1 is characterized in that: The step of training the LSTM-AdaBoost neural network model based on historical data specifically includes: Based on the AdaBoost ensemble algorithm, multiple LSTM networks are trained serially as base learners; Clean and normalize the historical data to obtain the training data set and the test data set; The LSTM-AdaBoost neural network model is trained. During the model training process, a three-layer LSTM network model is used, each layer contains 128 units, and the mean square error is used as the loss function. The Adam optimizer is used for optimization. In order to prevent the model from overfitting, the early stopping method and Dropout technology are used during the training process.

3. The customer carbon footprint prediction method based on power supply and demand forecasting and power trading contracts according to claim 2 is characterized in that: When training multiple LSTM networks as base learners, the training process starts with initializing the training set weights and setting the number of base learners T. Then the training sets are input into the LSTM network one by one for training. After each round of training, the relative error eti of the sample, the regression error rate εt of the LSTM network, and the weight coefficient αt of the base learner are calculated, and the weight distribution of the training set is updated. The above steps are repeated until the preset number of base learners T is reached. All trained base learners are linearly combined through their weight coefficients to form an LSTM-AdaBoost neural network model.

4. The customer carbon footprint prediction method based on power supply and demand prediction and power trading contract according to claim 2 is characterized in that: The steps of training the LSTM-AdaBoost neural network model specifically include: Initialize the weight vector, and the weight distribution of the training data is initialized as: For k = 1, 2, ..., K: When the weight distribution is D k When , the training data is trained through the LSTM neural network to obtain the predictor based on the LSTM model: h k :S→Y; Calculate h k Prediction error on the training dataset: Where M = sup i (|h k (x i )-y i |), the output interval of the function is [0,1]; calculate the total error of the training sample set: Calculate the current predictor h k The coefficient of in Update the weight distribution of the training dataset: in The loop ends and the connection weights of the LSTM predictor are recorded as W = (w1, w2, ..., w K ),in Construct the final predictor and integrate the trained predictors according to the connection weights to obtain the final strong predictor: h(x)=w1h1(x)+w2h2(x)+…+w K h K (x).

5. The customer carbon footprint prediction method based on power supply and demand prediction and power trading contract according to claim 1 is characterized in that: The step of matching the power generation side output with the power distribution customer load specifically includes: Based on the contract matrix, a constraint matrix form of supply and demand matching is constructed. The matrix P represents all power transaction relationships of supply and demand matching, where the rows represent power demand parties and the columns represent power suppliers. Based on the total demand, total supply and sources of power shortage, constraints are determined to match the output of the power generation side with the load of the power distribution customers.

6. The customer carbon footprint prediction method based on power supply and demand prediction and power trading contract according to claim 1 is characterized in that: The steps of predicting the future carbon footprint change trend of power distribution customers based on the carbon emission factor data of thermal power units specifically include: Based on the carbon emission factor data of thermal power units, combined with the real-time matching of power generation output and demand side load, the user's carbon footprint is calculated: The carbon emissions of various types of electricity are summarized to predict the carbon footprint of power distribution and power consumption customers. The calculation formula is: Total carbon footprint = ∑ electricity i × EF i , where i represents different types of electricity, the amount of electricity i is the amount of electricity consumed by the corresponding type, and EF i is the corresponding emission factor.

7. The customer carbon footprint prediction method based on power supply and demand prediction and power trading contract according to claim 6 is characterized in that: The electricity types include photovoltaic power, wind power and thermal power. For photovoltaic power and wind power, the emission factor is 0; for thermal power, EF is the real-time emission factor of the thermal power unit.

Images