A method for regulating demand-side response scheduling invitations that considers maximum demand.

Abstract

This invention relates to the field of virtual power plant participation in load resource demand response technology, and discloses an adjustment method for demand-side response scheduling invitations that considers maximum demand. The method includes the following steps: acquiring user load forecast impact data and inputting it into the user load forecast model to obtain the user's electricity load forecast curve; collecting the user's energy storage equipment information and the proposed demand response invitation information; calculating the user's maximum demand based on the user's electricity load forecast curve and the user's actual electricity load curve, calculating the output curve of the user's energy storage equipment during the response period, and then calculating the user's response capability curve, thereby adjusting the demand response invitation information, and submitting a response application based on the adjusted demand response invitation; during the invitation response period, monitoring the user's actual electricity load in real time, and adjusting the output of the energy storage equipment to ensure that the maximum demand limit is not exceeded; thus solving the technical problems of low efficiency in virtual power plant response invitations and poor ability to balance load fluctuations on the generation side.

Description

Technical Field

[0001] This invention relates to the field of virtual power plant participation in load resource demand response technology, and specifically to a regulation method that considers the demand-side response scheduling invitation based on maximum demand. Background Technology

[0002] With the transformation of the global energy structure, the installed capacity of new energy sources is increasing, such as the large-scale integration of solar and wind power. These new energy sources are characterized by intermittency and volatility. This instability leads to increased volatility in traditional power systems, posing a significant challenge to the safe and stable operation of the power grid.

[0003] To enhance the flexibility of new power systems and fully leverage the frequency regulation and peak shaving capabilities of user-side resources, the concept of virtual power plants has emerged. Virtual power plants integrate and coordinate distributed energy resources on the user side, giving them regulation capabilities similar to traditional power plants. This effectively balances load fluctuations on the generation side, improving the stability and reliability of the power system.

[0004] Demand-side response (DSR) refers to the behavior of electricity users who proactively change their conventional electricity consumption patterns in response to market price signals or incentive mechanisms. Essentially, virtual power plants achieve the goal of balancing fluctuations on the generation side through DSR, while electricity users also benefit economically. On the one hand, users can participate in DSR to consume electricity when prices are low or change their consumption behavior under the incentive of response subsidies, thereby reducing electricity costs or gaining additional revenue. On the other hand, the power grid can improve the operational efficiency and stability of the power system through the regulatory role of virtual power plants, reducing the need for reserve capacity and lowering operating costs.

[0005] However, current virtual power plant response invitation methods often fail to adequately consider the impact of response invitations on the electricity demand of the executing users. Since virtual power plants are primarily established to address grid-based issues, in practice, there is an overemphasis on response invitation efficiency to fully utilize the user load capacity.

[0006] In this model, users may significantly increase or decrease their electricity load when responding to invitations, leading to changes in user electricity demand. These changes in demand can have a substantial impact on user electricity bills. However, in many cases, the revenue generated from responding to invitations is far from sufficient to offset the increased electricity costs due to increased user demand.

[0007] In this situation, users cannot gain economic benefits from responding to demand response invitations and may even face economic losses due to participating in them. This significantly impacts users' willingness to respond to dispatch invitations, leading to a decrease in the number of users unwilling to participate in demand response invitations, thus reducing the participating group. Furthermore, the reduced load capacity of users participating in peak shaving and frequency regulation diminishes the effectiveness of virtual power plants, preventing them from fully leveraging their role in balancing load fluctuations on the generation side and improving the power system's flexible adjustment capabilities, creating a vicious cycle. Summary of the Invention

[0008] The present invention aims to provide a method for adjusting the demand-side response scheduling invitation that takes into account the maximum demand, so as to solve the technical problems of low efficiency of virtual power plant response invitation and poor ability to balance load fluctuations on the generation side.

[0009] To achieve the above objectives, the present invention adopts the following technical solution: a method for adjusting demand-side response scheduling invitations considering maximum demand, comprising the following steps:

[0010] Step 1: Obtain user load forecast impact data;

[0011] The second step is to input the user load forecast impact data into the user load forecast model to obtain the user's electricity load forecast curve for the next two days.

[0012] The third step is to collect the energy storage equipment information of the user who signed the contract based on the user ID of the demand-side response.

[0013] The fourth step is for the power grid dispatch center to obtain the proposed demand response invitation information;

[0014] Step 5: Calculate the user's maximum demand based on the user's predicted electricity load curve and the user's actual electricity load curve;

[0015] Step 6: Calculate the output curve of the user's energy storage equipment during the response period based on the demand response invitation information and the maximum demand, and then calculate the user's response capability curve during the response period. Adjust the demand response invitation information based on the user's energy storage equipment output curve and the user's response capability curve, and submit a response application based on the adjusted demand response invitation.

[0016] Step 7: During the invitation and response period, monitor the actual electricity load of users in real time, adjust the output of energy storage equipment to ensure that the maximum demand limit is not exceeded, and complete the demand-side response invitation.

[0017] The advantages of this solution are:

[0018] I. Break with conventional thinking and protect users' economic interests

[0019] Those skilled in the art often, due to the prevailing mindset that virtual power plants prioritize grid interests, tend to increase user-side electricity consumption or raise electricity prices to attract users when demand response efficiency is low. However, this approach often fails to truly improve users' economic benefits. For example, when implementing demand response in a certain region, simply raising the incentive price leads users to increase their electricity load to obtain subsidies, but the resulting increase in demand costs far exceeds the incentive benefits, thus increasing the actual economic burden on users.

[0020] This solution breaks away from this conventional thinking, fully considering the economic interests of users while ensuring the grid's regulation needs. Through reasonable planning and control, it ensures that users do not bear excessive additional costs due to increased electricity demand when participating in demand response, truly achieving a win-win situation for both users and the grid.

[0021] II. Balancing user electricity demand with grid regulation

[0022] Those skilled in the art generally believe that considering user electricity demand limits user load regulation capabilities during response invitations, affecting response invitation efficiency, and thus viewing it as contradictory to improving grid regulation capabilities. However, in reality, without considering user electricity demand, users may refuse to participate in the response due to concerns about increased electricity costs, thereby reducing grid regulation capabilities.

[0023] This solution cleverly strikes a balance between the two. Through precise calculations and rational planning, it fully leverages users' load regulation capabilities while meeting their electricity needs. For example, based on users' historical electricity consumption data and real-time load conditions, personalized response strategies are developed to both ensure normal electricity supply for users and provide effective regulation resources for the power grid.

[0024] III. Advance planning to resolve scheduling challenges

[0025] Those skilled in the art believe that dispatchers can only hastily formulate scheduling strategies that meet maximum demand limits based on real-time and uncertain load conditions at the moment of invitation and response. The decision-making time is tight and faces many uncertainties, making it difficult to achieve effective demand-limited scheduling.

[0026] This solution provides dispatchers with ample time for planning by forecasting user load conditions day-ahead. Dispatchers can plan the charging and discharging curves of flexible loads or energy storage devices during the invitation period in advance, much like drawing a "battle map" in advance. During the invitation response period, only minor adjustments need to be made based on the actual situation, without having to start from scratch, which greatly reduces the pressure and difficulty of real-time decision-making and enables the effective implementation of dispatching strategies that consider the maximum user demand.

[0027] IV. Demand control enhances user participation.

[0028] Through long-term work experience, those skilled in the art have creatively discovered the impact of demand control on the efficiency of the invitation effect. If demand is not controlled, exceeding the maximum demand will bring additional costs to responding users, causing them to be less proactive in responding to invitations, or even to completely stop participating in demand-side response, greatly reducing the number of participants in the invitation response group.

[0029] This solution eliminates users' concerns by strictly controlling maximum demand. Users no longer need to worry about high electricity bills due to exceeding demand limits when participating in demand response, thus encouraging them to actively respond to invitations. This not only enhances the grid's demand-side response dispatching capabilities but also promotes the healthy development of the demand-side response market.

[0030] V. Advance planning reduces the pressure of real-time decision-making.

[0031] By forecasting user load conditions in advance, dispatchers can plan the charging and discharging curves of flexible loads or energy storage devices during the invitation period. During the invitation response period, dispatchers only need to execute the pre-established plan and make minor adjustments based on real-time load conditions. This advance planning approach avoids the situation of hasty decisions at the moment of invitation response, greatly reducing the pressure and difficulty of real-time decision-making, and making dispatching work more efficient and orderly.

[0032] This solution, by comprehensively considering the maximum demand limit of users and optimizing the response invitation strategy, increases the enthusiasm of users to participate in the response invitation, thereby increasing the number of participants in the demand response invitation and enabling the power grid to dispatch more user load resources, effectively improving the effectiveness of the response invitation.

[0033] By fully leveraging the frequency regulation and peak shaving capabilities of user-side resources, this solution can balance load fluctuations on the generation side, improving the stability and reliability of the power system. When responding to sudden changes in electricity demand or fluctuations in renewable energy generation, the power grid can adjust more readily, ensuring the continuity and stability of power supply.

[0034] With the increasing integration of renewable energy sources, the grid faces growing pressure to absorb them. This solution, through demand-side response, guides users to increase electricity consumption when renewable energy generation is sufficient and decrease it when generation is insufficient. This helps promote the absorption and utilization of renewable energy, and drives the optimization and upgrading of the energy structure. For grid companies, this solution reduces the need for reserve capacity and lowers operating costs by optimizing dispatch strategies. Simultaneously, user participation in demand-side response can also reduce their own electricity costs to some extent, achieving a win-win situation for both the grid and users.

[0035] Preferably, as an improvement, the energy storage device information includes energy storage capacity, charging power, and discharging power; the demand response invitation information includes response type, response start time, response end time, and user load baseline during the response period; the response type includes valley filling response and peak shaving response;

[0036] The constraint equations are used to calculate the power output of the user's energy storage device during the response period. When the response type is valley filling response, the energy storage device needs to be charged during this period. The constraint equations are as follows:

[0037] ;

[0038] When the response type is peak-shaving response, the energy storage device needs to discharge during this period, and its constraint equations are as follows:

[0039] ;

[0040] in, This refers to the charging or discharging power of the energy storage device at that moment. This refers to the maximum charging power or maximum discharging power of the energy storage device. This represents the predicted user electricity load at that moment. Q represents the user's maximum demand, and Q represents the total capacity of the energy storage equipment.

[0041] The benefits of this improvement are as follows: By considering specific parameters of energy storage devices, such as storage capacity, charging power, and discharging power, the actual capacity of energy storage devices in demand response can be assessed more accurately. Furthermore, because the ramp-up rate of energy storage charging and discharging is relatively fast, rising from 0 to the maximum in no more than one minute, the ramp-up rate is not considered when calculating energy storage charging and discharging, simplifying the algorithm and reducing implementation difficulty. Different constraints are formulated according to different response types, such as valley filling response and peak shaving response, making the response invitation more aligned with the actual needs of the power grid and improving the relevance and effectiveness of the response.

[0042] Preferably, as an improvement, the user's response capability curve is calculated based on the energy storage device's output curve and the user's electricity load forecast curve. The calculation formula is as follows:

[0043] ;

[0044] in, For the user's responsiveness at that moment, This represents the predicted user electricity load at that moment. This represents the output power of the energy storage device at that moment. This represents the user's baseline power load at that moment.

[0045] The beneficial effects of this improvement are: by comprehensively considering the actual output capacity of energy storage devices and the predicted electricity load of users, the actual response capability of users in demand response can be assessed more accurately. This assessment method takes into account the dynamic characteristics of energy storage devices and the diversity of user electricity consumption behavior, making the assessment results closer to reality. Based on the calculation results of user response capability curves, the power grid dispatch center can formulate more optimized demand response strategies. For example, during peak load periods, users with strong response capabilities and high energy storage output can be prioritized to participate in peak shaving response, thereby more effectively balancing the power grid load.

[0046] Preferably, as an improvement, the user load forecast impact data includes the user's historical electricity consumption data, weather data of the user's location, and holiday data. The weather data includes historical temperature, future forecast temperature, historical humidity, and future forecast humidity; the holiday data includes the dates of historical and future holidays.

[0047] The beneficial effects of this improvement are as follows: historical electricity consumption data provides basic information on users' electricity consumption habits; weather data, such as temperature and humidity, reflects the impact of the external environment on users' electricity demand; and holiday data captures changes in users' electricity consumption patterns on special dates. Taking all these factors into account can significantly improve the accuracy of user load forecasting. Introducing multiple types of data allows the model to better adapt to changes in electricity demand from different users, regions, and time periods, improving the model's generalization ability and adaptability.

[0048] Preferably, as an improvement, a deep autoregressive recurrent neural network is used to construct the user load prediction model. The input layer of the user load prediction model is used to receive the user's historical 7-day electricity load data, the local historical 7-day weather data, the local weather forecast data for the next 2 days, and the local historical 7-day and next 2-day holiday information.

[0049] The hidden layer of the user load prediction model utilizes two recurrent neural network layers, with 40 neurons in each layer, and uses an activation function for nonlinear transformation.

[0050] The output layer of the user load forecasting model is used to output the predicted electricity load curves for users over the next two days.

[0051] The beneficial effect of this improvement is that deep autoregressive recurrent neural networks excel at processing time-series data and can capture the changing patterns of user electricity load over time. By introducing recurrent neural network layers, the model can better understand the relationship between historical data and future predictions.

[0052] The design of two recurrent neural network layers and 40 neurons per layer increases the depth and breadth of the model, enabling it to learn more complex nonlinear relationships and improving its expressive power. By outputting electricity load forecast curves for the next two days, the model can provide longer-term forecast information for grid dispatching and energy storage planning, helping to formulate more reasonable demand response strategies.

[0053] Preferably, as an improvement, the user load prediction model is set with a step size of 96*2 and a hyperparameter of 0.1 for random inactivation.

[0054] The beneficial effects of this improvement are: a step size of 962 means the model considers a longer time window during prediction. Assuming each step represents a certain time interval, such as 15 minutes, then 96 steps represent one day, and 962 steps represent two days. This helps the model capture electricity load change trends over a longer period, improving prediction stability. Random deactivation is a regularization technique that prevents overfitting by randomly discarding a portion of neurons during training. Setting the random deactivation rate to 0.1 means that 10% of neurons are randomly discarded in each training iteration. This helps reduce model complexity and improve the model's generalization ability.

[0055] Preferably, as an improvement, in the fifth step, the maximum value of the average power every 15 minutes in the user's monthly electricity consumption history data is taken as the user's maximum demand. The generated user electricity load prediction curve is compared with the maximum demand value to determine whether the value of the prediction curve exceeds the maximum demand value at any point in time. If there is a breakthrough point, the largest prediction value in the breakthrough point is taken as the new user's maximum demand.

[0056] The beneficial effects of this improvement are as follows: By directly using the maximum average power every 15 minutes from the user's current month's electricity consumption history data as the user's maximum demand, rather than based on the average of partial months or historical data, it can more accurately reflect the user's current electricity consumption habits and load characteristics. When a breakout point is predicted, the largest predicted value at the breakout point is used as the new user's maximum demand, which helps to formulate more effective demand response strategies. The power grid dispatch center can adjust the charging and discharging plans of energy storage devices and adjust users' electricity consumption plans in advance based on the new maximum demand forecast value to ensure that the user's electricity load does not exceed this new forecast value during the demand response period, thereby ensuring the stable operation of the power grid and the economic interests of users. Attached Figure Description

[0057] Figure 1 This is a flowchart of an embodiment of the present invention.

[0058] Figure 2 This is a bar chart showing the time taken to calculate the virtual power plant response in this invention. Detailed Implementation

[0059] The following detailed description illustrates the specific implementation method:

[0060] Example

[0061] The basics are as follows: Figure 1 As shown, a method for adjusting demand-side response scheduling invitations that considers maximum demand includes:

[0062] The first step is to obtain user load forecast impact data. This data includes historical electricity consumption data, weather data for the user's location, and holiday data. Weather data includes historical temperature, predicted future temperature, historical humidity, and predicted future humidity. Holiday data includes the dates of historical and future holidays.

[0063] The second step is to input the user load forecast impact data into the user load forecast model to obtain the user's electricity load forecast curve for the next two days.

[0064] A deep autoregressive recurrent neural network (DAR-RNN) is used to model the user load forecasting model. DAR-RNN is a model that combines the advantages of deep learning and recurrent neural networks (RNNs), and is particularly suitable for processing time series data.

[0065] The model was trained on the load prediction impact data of each demand-side response contracted user with a step size of 96*2, 2 layers, 40 hidden layers, and a random deactivation rate of 0.1. This setup divides a day into 96 time steps (e.g., every 15 minutes) and predicts load data for the next two days. The user load prediction model contains two recurrent neural network layers, each with 40 neurons in its hidden layer. The 0.1 hyperparameter is used to prevent overfitting by randomly discarding some neuron outputs during training.

[0066] The input layer of this user load forecasting model receives the user's historical 7-day electricity load data, the local area's historical 7-day weather data, the local area's weather forecast data for the next 2 days, and the local area's historical 7-day and next 2-day holiday information; the hidden layer uses two recurrent neural network layers, with 40 neurons in each layer, and uses an activation function (such as ReLU) for nonlinear transformation; the output layer of the user load forecasting model outputs the predicted user's electricity load forecast curve for the next 2 days.

[0067] The third step involves collecting the energy storage device information for the user who signed the demand-side response contract, based on their ID. This information includes energy storage capacity, charging power, and discharging power. Because the ramp-up rate of energy storage charging and discharging is relatively fast, rising from 0 to the maximum in no more than one minute, the ramp-up rate is not considered when calculating energy storage charging and discharging speeds to simplify the algorithm.

[0068] The fourth step involves the power grid dispatch center obtaining the demand response invitation information. This information includes the response type, response start time, response end time, and user load baseline for the response period. Response types include valley filling response and peak shaving response. The user load baseline for the response period represents the electricity load of users during the demand response period when they are not participating in the response.

[0069] Step 5: Calculate the user's maximum demand based on the user's predicted electricity load curve and the user's actual electricity load curve.

[0070] The user's maximum demand is the maximum average power consumption per 15 minutes within a certain billing cycle. Since the end of the month has not yet arrived, the user's actual maximum demand cannot be directly calculated. That is, the maximum demand on the electricity bill is calculated based on the user's maximum average power consumption per 15 minutes for this month. Therefore, a temporary maximum demand needs to be calculated as a reference based on the user's historical electricity consumption for this month. This value represents the user's maximum average power consumption per 15 minutes so far this month.

[0071] The generated user electricity load forecast curve is compared with the temporary maximum demand value to determine whether the forecast curve exceeds the temporary maximum demand value at any point in time. If there is a breakout point, the forecast value at the breakout point is taken as the new maximum demand. If the forecast curve exceeds the temporary maximum demand at multiple points in time, the largest value among them needs to be selected as the new maximum demand.

[0072] Step 6: Calculate the output of the user's energy storage equipment during the response period based on the demand response invitation information and the maximum demand, and calculate the user's invitation response capability at that moment. Adjust the demand response invitation based on the invitation response capability and the output of the user's energy storage equipment, and submit a response application based on the adjusted demand response invitation. The demand response invitation includes the invitation content of the user's responsiveness.

[0073] When the response type is valley filling response, the energy storage device needs to be charged during this period, and its constraint equations are as follows:

[0074] ;

[0075] When the response type is peak-shaving response, the energy storage device needs to discharge during this period, and its constraint equations are as follows:

[0076] ;

[0077] in, This refers to the charging or discharging power of the energy storage device at that moment. This refers to the maximum charging power or maximum discharging power of the energy storage device. This represents the predicted user electricity load at that moment. Q represents the user's maximum demand, and Q represents the total capacity of the energy storage equipment.

[0078] By solving the system of linear equations, the output power of the energy storage device at each moment can be obtained, i.e. Then, based on the energy storage device's output curve and the user's electricity load forecast curve, the user's response capability curve is calculated using the following formula:

[0079] ;

[0080] in, For the user's responsiveness at that moment, This represents the predicted user electricity load at that moment. This represents the output power of the energy storage device at that moment. This represents the user's baseline power load at that moment.

[0081] Step 7: During the invitation and response period, monitor the actual electricity load of users in real time, adjust the output of energy storage equipment to ensure that the maximum demand limit is not exceeded, and complete the demand-side response invitation.

[0082] This real-time monitoring and flexible control approach not only effectively controls electricity costs for users, preventing extra expenses due to over-demand, but also improves energy efficiency, achieving energy conservation and emission reduction. For power grid companies, it helps maintain stable grid operation, reduces grid load fluctuations, and improves the reliability and security of power supply.

[0083] This embodiment provides dispatchers with ample time for planning by forecasting user load conditions day-ahead. Dispatchers can plan the charging and discharging curves of flexible loads or energy storage devices during the invitation period in advance, much like drawing a "battle map" in advance. During the invitation response period, only minor adjustments need to be made according to the actual situation, without having to start from scratch, greatly reducing the pressure and difficulty of real-time decision-making.

[0084] As attached Figure 2As shown, a bar chart illustrating the computational time required for virtual power plant response calculations is provided to demonstrate the computational efficiency advantage of this solution. The computational time increases with the number of energy storage devices. However, comparing the results of manual calculations and those of this embodiment, it is clear that this embodiment is more efficient in terms of computational time. For example, when there is only one energy storage device, manual calculation takes 7 minutes, while this solution only takes 3.5 minutes; when the number of energy storage devices increases to three, manual calculation takes 22 minutes, while this embodiment only takes 4 minutes. This intuitively demonstrates the computational efficiency advantage of this solution, showing that even with more energy storage devices and more complex computational scenarios, this solution maintains a relatively high computational speed. This helps to make decisions faster in practical applications, improving the response speed and overall operational efficiency of the virtual power plant.

[0085] This technical solution not only breaks with conventional thinking and safeguards the economic interests of users, but also takes into account both user electricity demand and grid regulation, and resolves dispatching challenges through advance planning. (Attached) Figure 2 The demonstration of the virtual power plant response calculation time increasing with the number of energy storage devices further proves the advantage of this solution in terms of computational efficiency, enabling the implementation of response demand adjustment that takes into account user electricity demand.

[0086] The above descriptions are merely embodiments of the present invention, and common knowledge such as specific technical solutions and / or characteristics are not described in detail here. It should be noted that those skilled in the art can make various modifications and improvements without departing from the technical solutions of the present invention, and these should also be considered within the scope of protection of the present invention. These modifications and improvements will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A method for adjusting demand-side response scheduling invitations that considers maximum demand, characterized in that, Includes the following steps: Step 1: Obtain user load forecast impact data; The second step is to input the user load forecast impact data into the user load forecast model to obtain the user's electricity load forecast curve for the next two days. The third step is to collect the energy storage equipment information of the user who signed the contract based on the user ID of the demand-side response. The fourth step is for the power grid dispatch center to obtain the proposed demand response invitation information; Step 5: Calculate the user's maximum demand based on the user's predicted electricity load curve and the user's actual electricity load curve; Step 6: Calculate the output curve of the user's energy storage equipment during the response period based on the demand response invitation information and the maximum demand, and then calculate the user's response capability curve during the response period. Adjust the demand response invitation information based on the user's energy storage equipment output curve and the user's response capability curve, and submit a response application based on the adjusted demand response invitation. Step 7: During the invitation and response period, monitor the actual electricity load of users in real time, adjust the output of energy storage equipment to ensure that the maximum demand limit is not exceeded, and complete the demand-side response invitation. The energy storage device information includes energy storage capacity, charging power, and discharging power; the demand response invitation information includes response type, response start time, response end time, and user load baseline during the response period; the response type includes valley filling response and peak shaving response; The constraint equations are used to calculate the power output of the user's energy storage device during the response period. When the response type is valley filling response, the energy storage device needs to be charged during this period. The constraint equations are as follows: ； When the response type is peak-shaving response, the energy storage device needs to discharge during this period, and its constraint equations are as follows: ； in, This refers to the charging or discharging power of the energy storage device at that moment. This refers to the maximum charging power or maximum discharging power of the energy storage device. This represents the predicted user electricity load at that moment. Q represents the user's maximum demand, and Q represents the total capacity of the energy storage equipment. Based on the output curve of the energy storage device and the user's electricity load forecast curve, the user's response capability curve is calculated using the following formula: ； in, For the user's responsiveness at that moment, This represents the predicted user electricity load at that moment. This represents the output power of the energy storage device at that moment. This represents the user's baseline power load at that moment.

2. The adjustment method for demand-side response scheduling invitations considering maximum demand as described in claim 1, characterized in that: User load forecast impact data includes users' historical electricity consumption data, weather data for the user's location, and holiday data. Weather data includes historical temperature, future forecast temperature, historical humidity, and future forecast humidity; holiday data includes the dates of historical and future holidays.

3. The adjustment method for demand-side response scheduling invitations considering maximum demand according to claim 1, characterized in that: The user load prediction model is constructed using a deep autoregressive recurrent neural network. The input layer of the user load prediction model is used to receive the user's historical 7-day electricity load data, the local historical 7-day weather data, the local weather forecast data for the next 2 days, and the local holiday information for the past 7 days and the next 2 days. The hidden layer of the user load prediction model utilizes two recurrent neural network layers, with 40 neurons in each layer, and uses an activation function for nonlinear transformation. The output layer of the user load forecasting model is used to output the predicted electricity load curves for users over the next two days.

4. The adjustment method for demand-side response scheduling invitations considering maximum demand as described in claim 3, characterized in that: The user load prediction model is set with a step size of 96*2 and a hyperparameter of 0.1 for random inactivation.

5. The adjustment method for demand-side response scheduling invitations considering maximum demand according to claim 4, characterized in that: In the fifth step, the maximum value of the average power every 15 minutes in the user's electricity consumption history data for the current month is taken as the user's maximum demand. The generated user electricity load prediction curve is compared with the maximum demand value to determine whether the value of the prediction curve exceeds the maximum demand value at any point in time. If there is a breakthrough point, the largest prediction value in the breakthrough point is taken as the new user's maximum demand.

Images