User daily load curve prediction method
Through the LSTM recurrent neural network, users use energy load curve prediction model, combined with photovoltaic power generation prediction, calculate the elastic coefficient of load price and delay coefficient, and generate the user's daily load prediction curve, solving the problem of insufficient prediction accuracy in traditional methods and achieving higher prediction accuracy and adaptability.
Patent Information
- Application Number
- CN202510416529.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-03
- Publication Date
- 2025-07-18
AI Technical Summary
The traditional load curve prediction method has insufficient prediction accuracy and adaptability, which affects the optimization scheduling of the power grid and the operation of the energy economy, and fails to effectively consider the impact of photovoltaic power generation volatility and dynamic changes in time-sharing electricity prices on user energy consumption behavior.
The LSTM recurrent neural network is used to train the user's energy load curve prediction model, combined with the photovoltaic power generation prediction model, by calculating the load price elastic coefficient and delay coefficient, adjusting the energy load curve to peel off the influence of photovoltaic power generation, superimposing the load delay fluctuation curve, and generating the user's daily load prediction curve.
It improves the accuracy and practicality of user daily load curve prediction, can better adapt to photovoltaic power generation volatility and time-sharing electricity price changes, reduce the impact of electricity price fluctuations on load, and improve grid scheduling efficiency.
Smart Images
Figure CN120341831A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of power system load forecasting, and particularly relates to a method for predicting the daily load curve of users. Background Art
[0002] The safety and stability of the power system are crucial for the urban development and construction and the living security of residents. The stability of the power system refers to the ability of the system to maintain stability in the face of load changes and external disturbances. This requires the power system to have a reasonable balance between supply and demand and a good dispatching control mechanism to ensure the stability of power supply. With the deepening of the market-oriented power purchase business, the prediction accuracy of the daily load curve of users has an increasing impact on the power purchase costs of entities such as power grid companies acting as agents for power purchase, direct power purchase by user power plants, and power purchase by power sales companies acting as agents. At the same time, the electricity consumption pattern on the user side is also undergoing significant changes due to multiple factors.
[0003] However, traditional load curve prediction methods usually rely on factors such as historical load data and weather conditions, and there are large deviations between the prediction results and the actual load, resulting in technical problems such as decreased prediction accuracy and insufficient adaptability, which affect the optimal dispatching of the power grid and the economic operation of energy. Summary of the Invention
[0004] The purpose of the present invention is to provide a method for predicting the daily load curve of users, which can comprehensively consider the impact of the volatility of photovoltaic power generation and the dynamic change of time-of-use electricity price on the energy consumption behavior of users, so as to improve the prediction accuracy and practicality.
[0005] To achieve the above purpose, the present invention is implemented by adopting the following technical solutions: According to one aspect of the present invention, there is provided a method for predicting the daily load curve of users, including the following steps: Obtain the temperature data and holiday data of the prediction day, input them into the trained user energy consumption load curve prediction model, and output the energy consumption load prediction curve; Obtain the sunlight data and wind speed data of the prediction day, input them into the trained photovoltaic power generation prediction model, and output the photovoltaic power generation prediction curve; Obtain the historical user load data and calculate the monthly average daily energy consumption load curve; Obtain the time-of-use electricity price data and calculate the monthly average daily electricity price curve and the electricity price fluctuation curve; Calculate the load-price elasticity coefficient of the user by using the monthly average daily energy consumption load curve and the monthly average daily electricity price curve; Perform correlation time-delay analysis on the monthly average daily energy consumption load curve and the monthly average daily electricity price curve by using the cross-correlation function, and measure the optimal time-delay coefficient; Obtain the load time-delay fluctuation curve at the optimal time-delay coefficient by using the load-price elasticity coefficient and the electricity price fluctuation curve; Overlay the predicted curve of the energy consumption load and the curve of the load time-delay fluctuation, and deduct the predicted curve of the photovoltaic power generation to obtain the predicted curve of the user's daily load.
[0006] The implementation of time-of-use electricity price can affect the load distribution of users. However, due to factors such as the user's production organization ability and the time required for the load adjustment process of production equipment, there is a negative correlation time-delay characteristic between the load and the time-of-use electricity price.
[0007] Adopting the above technical solution, in view of the monthly change of the current user's time-of-use electricity price, the load-price elasticity coefficient of the user is calculated through the change amount of the average daily load and the change amount of the electricity price in adjacent months of the user, so as to quantify the change of the user's load fluctuation amount with the electricity price fluctuation at different months and different moments; after adjusting the predicted curve of the user's energy consumption load with the time-delay coefficient, the prediction accuracy can be improved. After adjusting the load-price elasticity coefficient and the time-delay of the predicted energy consumption load curve, and deducting the predicted photovoltaic power generation value, it is the predicted curve of the user's daily load.
[0008] According to an embodiment of the present invention, the prediction model of the user's energy consumption load curve is obtained by training with a long short-term memory (LSTM) recurrent neural network; During training, the input data includes the user's energy consumption load curve, the historical temperature curve, and the historical holiday curve, and the output data is the predicted curve of the energy consumption load.
[0009] Furthermore, the user's energy consumption load curve is obtained by superimposing the historical energy consumption load curve and the historical photovoltaic power generation curve.
[0010] Using the photovoltaic power generation power and the user's net load data at the same moment to restore the user's energy consumption load curve, after peeling off the influence of photovoltaic power generation on the user's load, an LSTM recurrent neural network is used to train the user's energy consumption load curve and its related influencing factors to generate an energy consumption load curve prediction model. In this way, the obtained predicted energy consumption load curve peels off the influence of the user-side photovoltaic power generation and can improve the prediction accuracy.
[0011] Using the photovoltaic power generation power and the user's net load data at the same moment to restore the user's energy consumption load curve, after peeling off the influence of photovoltaic power generation on the user's load, an LSTM recurrent neural network is used to train the user's energy consumption load curve and its related influencing factors to generate an energy consumption load curve prediction model. In this way, the obtained predicted energy consumption load curve peels off the influence of the user-side photovoltaic power generation and can improve the prediction accuracy.
[0012] When using an LSTM recurrent neural network to train the prediction model of the user's energy consumption load curve, the input data includes the user's energy consumption load curve 、the historical temperature curve and the historical holiday curve When obtaining data, select N moments per day for data collection.
[0013] Specifically, the energy consumption load curve is obtained by superimposing the historical energy consumption load curve and the historical photovoltaic power generation curve; among them, the historical energy consumption load curve is obtained from the historical load curve set, and the historical photovoltaic power generation curve is obtained from the historical photovoltaic power generation curve set.
[0014] The energy consumption load curve is calculated according to the following formula: , In the formula, M represents the month, and D represents the date; represents the historical load curve set on the Dth day of the Mth month, represents the historical photovoltaic power generation curve set on the Dth day of the Mth month; represents the energy consumption load curve on the yth day of the xth month.
[0015] The historical load curve set is as follows: , represents the load curve on the yth day of the xth month; The historical photovoltaic power generation curve set is as follows: , represents the photovoltaic power generation curve on the yth day of the xth month; The energy consumption load curve on the yth day of the xth month is calculated according to the following formula:
[0016] In the formula, ; N represents the number of moments for data collection per day.
[0017] The historical temperature data is obtained by the following formula: , represents the temperature curve on the yth day of the xth month.
[0018] The holiday data is obtained by the following formula: , represents the holiday type on the yth day of the xth month; specifically, when it is a working day, when it is an ordinary holiday (weekend), and when it is a special holiday (Spring Festival, Tomb-Sweeping Festival, Labor Day, National Day, etc.).
[0019] Key steps and parameter formulas of the LSTM recurrent neural network:
[0020]
[0021]
[0022]
[0023]
[0024] wherein: i and f and c and o are the input gate, forget gate, cell state, and output gate respectively: W and b are the corresponding weight coefficient matrix and bias term respectively; σ and tanh are the sigmoid and hyperbolic tangent activation functions respectively.
[0025] According to an embodiment of the present invention, the photovoltaic power prediction model is obtained by training a long short-term memory recurrent neural network; During training, the input data includes the historical photovoltaic power curve, historical light intensity curve, and historical wind speed curve.
[0026] The input training data of the LSTM recurrent neural network includes the historical photovoltaic power curve set the historical light intensity curve set the wind speed curve set ; The historical photovoltaic power curve set is obtained according to the following formula: wherein, represents the photovoltaic power curve on the yth day of the xth month; The historical light intensity curve set is obtained according to the following formula: wherein, represents the light intensity curve on the yth day of the xth month; The historical wind speed curve set is obtained according to the following formula: wherein, represents the wind speed curve on the yth day of the xth month.
[0027] According to an embodiment of the present invention, in the step of calculating the load price elasticity coefficient of the user by using the monthly average daily load curve and the monthly average daily electricity price curve, the load price elasticity coefficient is calculated by the following formula: , wherein, N represents the number of data collection times per day; represents the monthly average daily energy consumption load curve, represents the monthly average daily electricity price curve; represents the load - electricity price elasticity coefficient at the n - th moment of each day in the M - th month.
[0028] The monthly average daily energy consumption load curve is obtained by using the historical energy consumption load curve ; The monthly average daily electricity price curve is obtained by using the historical electricity price curve .
[0029] Specifically, N moments of each day are selected for data collection, The monthly average daily energy consumption load curve is obtained according to the following formula: , where N represents the number of moments for data collection each day; represents the user load data collected at the n - th moment of the d - th day in the M - th month; represents the average value of the user load data collected at the n - th moment of each day in the M - th month.
[0030] The monthly average daily electricity price curve is obtained according to the following formula: , where N represents the number of moments for data collection each day; represents the electricity price data collected at the n - th moment of the d - th day in the M - th month; represents the average value of the electricity price data collected at the n - th moment of each day in the M - th month.
[0031] , , represents the electricity price curve of the y - th day in the x - th month, represents the electricity price data collected at the n - th moment of the y - th day in the x - th month.
[0032] According to an embodiment of the present invention, in the step of performing correlation time - delay analysis on the monthly average daily energy consumption load curve and the monthly average daily electricity price curve by using the cross - correlation function and calculating the optimal time - delay coefficient, the following steps are included: Obtain historical user load data, calculate the monthly average daily energy consumption load curve, and calculate the Q order difference curve of the monthly average daily energy consumption load curve; Obtain time - of - use electricity price data, calculate the monthly average daily electricity price curve, and calculate the Q order difference curve of the monthly average daily electricity price curve; Calculating the monthly average daily energy consumption load curve Q The Q-order difference curve of the monthly average daily energy consumption load curve and the monthly average daily electricity price curve Q The correlation coefficient of the Q-order difference curve at all time delay coefficients; Select the optimal time delay coefficient according to the obtained correlation coefficient.
[0033] Among them, the parameter , N represents the number of data collection times per day.
[0034] Specifically, select N times per day for data collection to obtain the monthly average daily energy consumption load curve and the average daily electricity price curve ; Calculate the Q-order difference curve of the monthly average daily energy consumption load curve according to the following formula : ; Calculate the Q-order difference curve of the average daily electricity price curve according to the following formula : ; The parameter Q is a positive integer, and 1 ≤ Q ≤ N, N represents the number of data collection times per day; Q The selection of Q can be defined according to the regional statistical characteristics; generally, The value of can be screened and confirmed by the following method: Traverse all when
[0035] The Q value corresponding to Q The Q-order difference curve of the monthly average daily energy consumption load curve and the monthly average daily electricity price curve Q The correlation coefficient of the Q-order difference curve at all time delay coefficients is calculated according to the following formula: ; In the formula, ; ; Traverse all the correlation coefficients when to obtain the optimal time delay coefficient.
[0036] Specifically, when it is considered that the user does not have the negative correlation time delay characteristic, and the optimal time delay coefficient is 0 at this time; when at KThe value is the optimal time delay coefficient.
[0037] Obtain the time-of-use electricity price data and calculate the electricity price fluctuation curve for the x-th month according to the following formula : ; In the formula, N represents the number of data collection times per day; represents the electricity price fluctuation value at the n-th moment of each day in the x-th month.
[0038] According to an embodiment of the present invention, the load time delay fluctuation curve under the optimal time delay coefficient is calculated according to the following formula: , In the formula, ; K is the optimal time delay coefficient; N represents the number of data collection times per day; represents the electricity price fluctuation curve for the x-th month.
[0039] According to an embodiment of the present invention, the user's daily load prediction curve is calculated according to the following formula: .
[0040] According to an aspect of the present invention, there is provided a user daily load prediction device, including: An energy consumption load prediction module, configured to obtain the predicted day temperature data and holiday data, input the trained user energy consumption load curve prediction model, and output the energy consumption load prediction curve; A photovoltaic power generation prediction module, configured to obtain the predicted day light data and wind speed data, input the trained photovoltaic power generation power prediction model, and output the photovoltaic power generation prediction curve; A time delay analysis module, configured to obtain historical user load data and calculate the monthly average daily energy consumption load curve; Obtain the time-of-use electricity price data, calculate the monthly average daily electricity price curve and the electricity price fluctuation curve; Calculate the load electricity price elasticity coefficient of the user by using the monthly average daily energy consumption load curve and the monthly average daily electricity price curve; Perform correlation time delay analysis on the monthly average daily energy consumption load curve and the monthly average daily electricity price curve by using the cross-correlation function to measure the optimal time delay coefficient; Obtain the load time delay fluctuation curve under the optimal time delay coefficient by using the load electricity price elasticity coefficient and the electricity price fluctuation curve; A prediction module, configured to superimpose the energy consumption load prediction curve and the load time delay fluctuation curve, and deduct the photovoltaic power generation prediction curve to obtain the user daily load prediction curve.
[0041] According to one aspect of the present invention, there is provided a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the user daily load curve prediction method of any of the above embodiments is implemented.
[0042] According to one aspect of the present invention, there is provided a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the user daily load curve prediction method considering the influence of photovoltaic power generation and time-of-use electricity price is implemented.
[0043] Compared with the prior art, the present invention has at least the following beneficial effects: 1. By using the photovoltaic power generation power and user net load data at the same time to restore the user's energy consumption load curve, after separating the influence of photovoltaic power generation on the user load, the LSTM neural network is used to train the energy consumption load curve and its related influencing factors to generate an energy consumption load curve prediction model, which can ensure the accuracy of the prediction result; 2. Aiming at the situation that the user's time-of-use electricity price changes monthly, the load-price elasticity coefficient of the user is calculated through the monthly average daily load change amount and electricity price change amount of adjacent months of the user to quantify the change of the user load fluctuation amount with the electricity price fluctuation at different months and different times; 3. The cross-correlation function is used to perform correlation time-delay analysis on the first-order difference curve of the daily load and the first-order difference curve of the electricity price, and the optimal time-delay coefficient is calculated, which can reduce or eliminate the influence of the time-of-use electricity price on the user load curve; 4. After adjusting the load-price elasticity coefficient and time-delay of the predicted energy consumption load curve, and deducting the predicted photovoltaic power generation power value, it is the predicted user daily load curve, which can greatly improve the prediction accuracy and is suitable for further popularization and application. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] The accompanying drawings forming a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings: Figure 1 It is a schematic diagram of the training method of the user energy consumption load curve prediction model in Embodiment 1 of the present invention; Figure 2 It is a schematic diagram of the training method of the photovoltaic power generation power prediction model in Embodiment 1 of the present invention; Figure 3 It is a schematic flowchart of the LSTM recurrent neural network algorithm in the training method of the user energy consumption load curve prediction model in Embodiment 1 of the present invention; Figure 4Schematic diagram of the method steps for obtaining the user's daily load prediction curve in Embodiment 1 of the present invention. Detailed implementation manners
[0045] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.
[0046] The following detailed descriptions are all exemplary descriptions, aiming to provide further detailed descriptions of the present invention. Unless otherwise specified, all technical terms adopted by the present invention have the same meaning as commonly understood by those of ordinary skill in the art to which the present invention belongs. The terms used in the present invention are only for describing specific implementation manners and are not intended to limit the exemplary embodiments according to the present invention.
[0047] Embodiment 1 A method for predicting the user's daily load curve, as Figures 1-4 shown, includes the following steps: Obtain the temperature data and holiday data of the prediction day, input them into the trained user energy consumption load curve prediction model, and output the energy consumption load prediction curve; Obtain the sunlight data and wind speed data of the prediction day, input them into the trained photovoltaic power generation prediction model, and output the photovoltaic power generation prediction curve; Obtain the historical user load data and calculate the monthly average daily energy consumption load curve; Obtain the time-of-use electricity price data and calculate the monthly average daily electricity price curve and the electricity price fluctuation curve; Calculate the load-price elasticity coefficient of the user by using the monthly average daily energy consumption load curve and the monthly average daily electricity price curve; Perform correlation time-delay analysis on the monthly average daily energy consumption load curve and the monthly average daily electricity price curve by using the cross-correlation function, and measure the optimal time-delay coefficient; Obtain the load time-delay fluctuation curve under the optimal time-delay coefficient by using the load-price elasticity coefficient and the electricity price fluctuation curve; Superimpose the energy consumption load prediction curve and the load time-delay fluctuation curve, and deduct the photovoltaic power generation prediction curve to obtain the user's daily load prediction curve.
[0048] Among them, the data is collected according to the method of collecting 96-point curves, that is, 96 moments are selected for data collection every day. The specific steps are as follows: S1. Restore the energy consumption load curve by using the photovoltaic power generation data and the user load data at the same moment .
[0049] The photovoltaic power generation data includes the historical photovoltaic power generation curve set , and the historical photovoltaic power generation curve set is obtained according to the following formula: , , Wherein, represents the photovoltaic power generation curve on the y-th day of the x-th month, represents the historical photovoltaic power generation data collected at the n-th moment on the y-th day of the x-th month.
[0050] The user load data includes a set of historical load curves , and the set of historical load curves is obtained according to the following formula: , , Wherein, represents the load curve on the y-th day of the x-th month.
[0051] The set of historical load curves can be obtained by the power grid company collecting the user's online meter data; when represents that the user draws electric energy from the power grid to meet its own energy consumption needs, represents that the user feeds electricity back to the power grid.
[0052] The energy consumption load curve restored by using the photovoltaic power generation data and the user load data at the same moment is: , , ; ; .
[0053] Wherein, M represents the month and D represents the date; represents the set of historical load curves, represents the set of historical photovoltaic power generation curves; represents the energy consumption load curve on the y-th day of the x-th month; represents the actual energy consumption load data of the user at the n-th moment on the y-th day of the x-th month; represents the historical energy consumption load data collected at the n-th moment on the y-th day of the x-th month; represents the historical photovoltaic power generation data collected at the n-th moment on the y-th day of the x-th month.
[0054] S2. Use the LSTM recurrent neural network to train the energy consumption load curve prediction model.
[0055] During training, the input data includes the user's energy consumption load curve, historical temperature curve, and historical holiday curve, and the output data is the energy consumption load prediction curve. Among them, the historical energy consumption load curve is obtained from the historical load curve set; the historical photovoltaic power generation curve is obtained from the historical photovoltaic power generation curve set.
[0056] The historical temperature data is obtained using the following formula: , , where: represents the temperature curve for the y-th day of the x-th month; represents the temperature data collected at the n-th moment on the y-th day of the x-th month.
[0057] The holiday data is obtained using the following formula: , where: represents the holiday type for the y-th day of the x-th month; specifically, when it is a working day, when it is a common holiday (weekend), and when it is a special holiday (such as Spring Festival, Tomb-Sweeping Day, Labor Day, National Day, etc.).
[0058] Take , and as the input training data for the LSTM recurrent neural network.
[0059] Key steps and parameter formulas of the LSTM recurrent neural network:
[0060]
[0061]
[0062]
[0063]
[0064] where: i , f , c , o are the input gate, forget gate, cell state, and output gate respectively: W and b are the corresponding weight coefficient matrices and bias terms respectively; σ and tanh are the sigmoid and hyperbolic tangent activation functions respectively.
[0065] S3. Use the LSTM recurrent neural network to train and generate a photovoltaic power prediction model.
[0066] During training, the input data includes the input training data of the LSTM recurrent neural network, including the set of historical photovoltaic power curves , the set of historical light intensity curves , and the set of wind speed curves ; the output data is the photovoltaic power prediction curve.
[0067] The set of historical photovoltaic power curves is obtained according to the following formula: , , where represents the photovoltaic power curve on the y-th day of the x-th month; represents the historical photovoltaic power data collected at the n-th moment on the y-th day of the x-th month.
[0068] The set of historical light intensity curves is obtained according to the following formula: , , where represents the light intensity curve on the y-th day of the x-th month; represents the historical light intensity data collected at the n-th moment on the y-th day of the x-th month.
[0069] The set of historical wind speed curves is obtained according to the following formula: , , where represents the wind speed curve on the y-th day of the x-th month; represents the historical light speed data collected at the n-th moment on the y-th day of the x-th month.
[0070] S4. Use the user's monthly average daily energy consumption load curve and the monthly average daily electricity price curve to calculate the user's load electricity price elasticity coefficient .
[0071] First, calculate the user's monthly average daily energy consumption load curve and the monthly average daily electricity price curve , where the monthly average daily energy consumption load curve Obtained based on historical user load data: , wherein, represents the user load data collected at the nth moment on the dth day of the Mth month; represents the average value of the user load data collected at the nth moment of each day in the Mth month.
[0072] The monthly average daily electricity price curve of the user is obtained based on time-of-use electricity price data; specifically,
[0073] wherein, represents the electricity price data collected at the nth moment on the dth day of the Mth month; represents the average value of the electricity price data collected at the nth moment of each day in the Mth month.
[0074] , , represents the electricity price curve on the yth day of the xth month, represents the electricity price data collected at the nth moment on the yth day of the xth month.
[0075] Load electricity price elasticity coefficient is calculated according to the following formula: , wherein, represents the monthly average daily energy consumption load curve, represents the monthly average daily electricity price curve; represents the load electricity price elasticity coefficient at the nth moment of each day in the Mth month.
[0076] S5. Calculate the negative correlation delay coefficient between the user's load and the time-of-use electricity price K ; Obtain the optimal delay coefficient.
[0077] S5-1 Calculate the monthly average daily energy consumption load curve of Q the -th order difference curve according to the following formula: Calculate the monthly average daily electricity price curve of Q the -th order difference curve according to the following formula: The parameter Q is a positive integer, and 1 ≤ Q≤N, where N represents the number of data collection times per day; Q The selection of Q can be defined according to the statistical characteristics of the region; generally, Q The value of Q can be screened and confirmed by the following method: Traverse all , The corresponding Q value is the optimal order.
[0078] S5-2 Calculate according to the following formula and the correlation coefficient K under the time delay coefficient : ; Among them, ; .
[0079] S5-3 Traverse all the correlation coefficients to obtain the optimal time delay coefficient.
[0080] Specifically, when , it is considered that the user does not have the negative correlation time delay characteristic, and the optimal time delay coefficient is 0 at this time; when the K value is the optimal time delay coefficient.
[0081] S6. Obtain the load time delay fluctuation curve under the optimal time delay coefficient.
[0082] First, calculate the electricity price fluctuation curve of the x-th month according to the electricity price: ; In the formula, represents the electricity price fluctuation value at the n-th moment of each day in the x-th month.
[0083] Then, calculate the load time delay fluctuation curve under the optimal time delay coefficient according to the following formula: ; In the formula, , K is the optimal time delay coefficient.
[0084] S7. Obtain the user's daily load prediction curve.
[0085] Obtain the predicted day's temperature data and holiday data, input them into the trained user energy load curve prediction model, and output the energy load prediction curve ; Obtain the predicted daily sunlight data and wind speed data, input them into the trained photovoltaic power generation prediction model, and output the photovoltaic power generation prediction curve ; The user daily load prediction curve is as follows: 。
[0086] Adopt the user daily load curve prediction method considering the influence of photovoltaic power generation and time-of-use electricity price provided in this embodiment. Use the photovoltaic power generation power and user net load data at the same time to restore the user's energy consumption load curve. After stripping the influence of photovoltaic power generation on the user load, use the LSTM neural network to train the energy consumption load curve and its related influencing factors to generate an energy consumption load curve prediction model to ensure the accuracy of the prediction results; use the cross-correlation function to conduct correlation time-delay analysis on the first-order difference curve of the daily load and the first-order difference curve of the electricity price to calculate the optimal time-delay coefficient, which can reduce or eliminate the influence of time-of-use electricity price on the user load curve; after adjusting the load electricity price elasticity coefficient and time-delay of the predicted energy consumption load curve and subtracting the predicted photovoltaic power generation power value, it is the predicted user daily load curve, which can greatly improve the prediction accuracy and is suitable for further popularization and application.
[0087] Embodiment 2 A user daily load prediction device includes: An energy consumption load prediction module, configured to obtain the predicted daily temperature data and holiday data, input them into the trained user energy consumption load curve prediction model, and output the energy consumption load prediction curve; A photovoltaic power generation prediction module, configured to obtain the predicted daily sunlight data and wind speed data, input them into the trained photovoltaic power generation prediction model, and output the photovoltaic power generation prediction curve; A time-delay analysis module, configured to obtain historical user load data and calculate the monthly average daily energy consumption load curve; obtain time-of-use electricity price data and calculate the monthly average daily electricity price curve and electricity price fluctuation curve; use the monthly average daily energy consumption load curve and the monthly average daily electricity price curve to calculate the load electricity price elasticity coefficient of the user; use the cross-correlation function to conduct correlation time-delay analysis on the monthly average daily energy consumption load curve and the monthly average daily electricity price curve to calculate the optimal time-delay coefficient; obtain the load time-delay fluctuation curve under the optimal time-delay coefficient by using the load electricity price elasticity coefficient and the electricity price fluctuation curve; A prediction module, configured to superimpose the energy consumption load prediction curve and the load time-delay fluctuation curve, and subtract the photovoltaic power generation prediction curve to obtain the user daily load prediction curve.
[0088] Embodiment 3 A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the user daily load curve prediction method of Embodiment 1.
[0089] Example 4 A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the user daily load curve prediction method considering photovoltaic power generation and the influence of time-of-use electricity price in Example 1.
[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications or equivalent substitutions can still be made to the specific embodiments of the present invention, and any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered by the protection scope of the claims of the present invention.
Claims
1. A method for predicting the daily load curve of a user, characterized in that, It includes the following steps: Obtain the temperature data and holiday data of the prediction day, input them into the trained user energy consumption load curve prediction model, and output the energy consumption load prediction curve; Obtain the sunlight data and wind speed data of the prediction day, input them into the trained photovoltaic power generation prediction model, and output the photovoltaic power generation prediction curve; Obtain the historical user load data and calculate the monthly average daily energy consumption load curve; Obtain the time-of-use electricity price data and calculate the monthly average daily electricity price curve and the electricity price fluctuation curve; Calculate the load electricity price elasticity coefficient of the user by using the monthly average daily energy consumption load curve and the monthly average daily electricity price curve; Perform correlation time-delay analysis on the monthly average daily energy consumption load curve and the monthly average daily electricity price curve by using the cross-correlation function, and measure the optimal time-delay coefficient; Obtain the load time-delay fluctuation curve under the optimal time-delay coefficient by using the load electricity price elasticity coefficient and the electricity price fluctuation curve; Superimpose the energy consumption load prediction curve and the load time-delay fluctuation curve, and deduct the photovoltaic power generation prediction curve to obtain the user daily load prediction curve.
2. The method for predicting the user daily load curve according to claim 1, wherein The user energy consumption load curve prediction model is obtained by training with a long short-term memory recurrent neural network; During training, the input data includes the user energy consumption load curve, the historical temperature curve, and the historical holiday curve, and the output data is the energy consumption load prediction curve.
3. The method for predicting the user daily load curve according to claim 2, wherein The user energy consumption load curve is obtained by superimposing the historical energy consumption load curve and the historical photovoltaic power generation curve.
4. The method for predicting the user daily load curve according to claim 1, wherein The photovoltaic power generation prediction model is obtained by training with a long short-term memory recurrent neural network; During training, the input data includes the historical photovoltaic power generation curve, the historical sunlight intensity curve, and the historical wind speed curve.
5. The method for predicting the user daily load curve according to claim 1, wherein In the step of calculating the load electricity price elasticity coefficient of the user by using the monthly average daily energy consumption load curve and the monthly average daily electricity price curve, The monthly average daily energy consumption load curve is obtained by using the historical energy consumption load curve; The monthly average daily electricity price curve is obtained by using the historical electricity price curve.
6. The method for predicting the user daily load curve according to claim 1, wherein In the step of performing correlation time-delay analysis on the monthly average daily energy consumption load curve and the monthly average daily electricity price curve by using the cross-correlation function and measuring the optimal time-delay coefficient, it includes the following steps: Obtain historical user load data, calculate the monthly average daily energy consumption load curve, and calculate the Q first-order difference curve; Obtain time-of-use electricity price data, calculate the monthly average daily electricity price curve, and calculate the Q first-order difference curve of the monthly average daily electricity price curve; Calculating the Q first-order difference curve of the monthly average daily energy consumption load curve and the Q first-order difference curve of the monthly average daily electricity price curve, and the correlation coefficient at all time delay coefficients; Screen the optimal time-delay coefficient according to the obtained correlation coefficient.
7. The method for predicting the user daily load curve according to claim 6, wherein In the step of screening the optimal time-delay coefficient according to the obtained correlation coefficient, When the minimum value of the obtained correlation coefficient is not less than -0.3, the optimal time-delay coefficient is 0; When the minimum value of the obtained correlation coefficient is less than -0.3, the corresponding time-delay coefficient is the optimal time-delay coefficient.
8. A user daily load prediction device, characterized in that, It includes: An energy consumption load prediction module, configured to obtain the temperature data and holiday data of the prediction day, input them into the trained user energy consumption load curve prediction model, and output the energy consumption load prediction curve; The photovoltaic power generation prediction module is used to obtain the predicted daily sunlight data and wind speed data, input them into the trained photovoltaic power generation prediction model, and output the photovoltaic power generation prediction curve; The time-delay analysis module is used to obtain the historical user load data and calculate the monthly average daily load curve of energy consumption; obtain the time-of-use electricity price data and calculate the monthly average daily electricity price curve and the electricity price fluctuation curve; calculate the load-price elasticity coefficient of the user by using the monthly average daily load curve of energy consumption and the monthly average daily electricity price curve; conduct a correlation time-delay analysis on the monthly average daily load curve of energy consumption and the monthly average daily electricity price curve by using the cross-correlation function to measure the optimal time-delay coefficient; obtain the load time-delay fluctuation curve under the optimal time-delay coefficient by using the load-price elasticity coefficient and the electricity price fluctuation curve; The prediction module is used to superimpose the energy consumption load prediction curve and the load time-delay fluctuation curve, and deduct the photovoltaic power generation prediction curve to obtain the user daily load prediction curve.
9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements the user daily load curve prediction method according to any one of claims 1-7 when executing the computer program.
10. A computer-readable storage medium storing a computer program, where the computer program implements the user daily load curve prediction method according to any one of claims 1-7 when executed by a processor.