A method, system, device and storage medium for optimizing power consumption

Abstract

This invention discloses a method, system, device, and storage medium for electricity consumption optimization. The method includes: predicting electricity prices using an electricity price prediction model based on electricity price-related data and external electricity consumption impact data for a target area, thereby obtaining electricity price prediction data; the electricity price-related data includes clearing price data and electricity price data for the target area within a set time period; the external electricity consumption impact data includes meteorological data, calendar characteristic data, and electricity supply-demand ratio for the target area within a set time period; performing cluster analysis on user electricity settlement data to construct a user load characteristic profile; and conducting electricity consumption simulation based on the electricity price prediction data and load characteristic profile, with the goal of minimizing electricity costs, to obtain the user's target electricity load data and generate the user's electricity consumption optimization strategy. This invention can achieve user-side load collaborative optimization and electricity dispatch optimization that integrates electricity price prediction and user electricity consumption behavior.

Description

Technical Field

[0001] This invention relates to the fields of data processing and artificial intelligence technology, and in particular to a power optimization method, system, device and storage medium. Background Technology

[0002] Against the backdrop of energy structure transformation, the market-oriented reform of the electricity market continues to deepen. The large-scale integration of new energy sources has led to high-frequency fluctuations in electricity prices, highlighting the pressure on users to control electricity costs. Currently, due to the difficulty in effectively transmitting spot time-of-use electricity price signals to end users, and the significant differences in load curves among different users (such as different daytime and nighttime electricity consumption ratios), there are huge differences in electricity costs, making it difficult for users to enjoy the benefits of marketization.

[0003] Currently, existing predictive frameworks that combine data and physical models first acquire boundary conditions such as grid topology, transmission constraints, and supply and demand information. They then construct simplified, security-constrained economic dispatch models and, by simulating the market clearing process, obtain theoretical real-time electricity price predictions. However, existing technologies only provide price predictions without analyzing actual user electricity consumption or identifying adjustable and non-adjustable loads. Therefore, they struggle to provide effective suggestions for optimizing electricity consumption strategies, resulting in a disconnect between price predictions and electricity consumption response. Summary of the Invention

[0004] To address the problems existing in the prior art, embodiments of the present invention provide a power consumption optimization method, system, device, and storage medium, which can achieve user-side load collaborative optimization and power consumption scheduling optimization by integrating electricity price forecasts and user power consumption behavior.

[0005] In a first aspect, embodiments of the present invention provide an electricity optimization method, comprising: Based on electricity price data and external electricity consumption impact data for the target area, a pre-trained electricity price prediction model is used to predict electricity prices, resulting in predicted electricity price data. The electricity price data includes: clearing price data and electricity price data for the target area within a specified time period; the external electricity consumption impact data includes: meteorological data, calendar characteristic data, and electricity supply-demand ratio for the target area within the specified time period. Cluster analysis is performed on the electricity billing data of each user in the target area to construct a load characteristic profile of the corresponding user; Based on the electricity price forecast data and the load characteristic profile, electricity consumption simulation is performed with the goal of minimizing electricity costs to obtain the target electricity load data for the corresponding user within a future set time. Based on the target electricity load data, an electricity consumption optimization strategy for the corresponding user within the future set time is generated.

[0006] As an improvement to the above scheme, based on electricity price data and external electricity consumption impact data for the target area, a pre-trained electricity price prediction model is used to predict electricity prices, resulting in electricity price prediction data, including: Data preprocessing is performed on the electricity price-related data and the external electricity consumption impact data; Correlation analysis was performed on the preprocessed electricity price-related data and external electricity consumption impact data to screen out electricity price-related features and external electricity consumption impact features whose correlation with electricity price meets the preset correlation conditions. The external electricity consumption impact features and the electricity price related features are time-aligned to construct an input feature vector; The input feature vector is fed into the electricity price prediction model to predict electricity prices, thereby obtaining electricity price prediction data.

[0007] As an improvement to the above scheme, cluster analysis is performed on the electricity billing data of each user in the target area to construct a load characteristic profile for the corresponding user, including: For each user's electricity billing data, the K-means clustering algorithm is used to extract the daily load curves for different daily scenarios from the electricity billing data; The daily load curve is decomposed to determine the rigid load and elastic load for each time period under the corresponding daily type scenario; Based on the rigid and flexible loads at different times under different day types, a load characteristic profile of the corresponding user under different day types is constructed.

[0008] As an improvement to the above scheme, the daily load curve is decomposed to determine the rigid and flexible loads for each time period under the corresponding daily scenario, including: For the daily load curves under different day types, a moving average is performed on the daily load curves to obtain the load baseline; The load baseline is filtered by quantiles to obtain the rigid load for each time period under the corresponding day type scenario; Based on the daily load curve and the rigid load of each time period, calculate the elastic load of each time period under the corresponding daily type scenario.

[0009] As an improvement to the above scheme, based on the electricity price forecast data and the load characteristic profile, an electricity consumption simulation is performed with the goal of minimizing electricity costs, resulting in the target electricity load data for the corresponding users within a set future time period, including: The first time interval corresponding to the predicted electricity price that is higher than the set first electricity price threshold is selected from the electricity price prediction data; Based on the load characteristic profile, calculate the transferred load amount for the first time interval; In a pre-built virtual settlement environment, the transferred load is transferred from the first time interval to the corresponding transferable time period, and a simulated settlement is performed according to the electricity price forecast data to obtain the final transferred load with the goal of minimizing electricity costs. Based on the final transferred load and the daily load curve, the target electricity billing data for the corresponding users within a set future time period is determined.

[0010] As an improvement to the above scheme, based on the load characteristic profile, the amount of transferred load in the first time interval is calculated, including: Extract the elastic load corresponding to the first time interval from the load characteristic profile, and use it as the upper limit of the transferable load for the first time interval; Based on the load characteristic profile, calculate the total load capacity of the elastic load transfer time window in the first time interval; The minimum value between the upper limit of transferable load and the total load capacity is selected as the amount of transferable load for the first time interval.

[0011] As an improvement to the above scheme, the calculation process of the elastic load transferable time window includes: Based on the time period of the rigid load, the time period of the elastic load, and the preset power consumption constraints, calculate the transferable time period of the elastic load in each time period; Based on the rigid and flexible loads for each time period, and the actual load for the corresponding time period in the daily load curve, calculate the transferable power range for each time period. The transferable time window for flexible loads is determined based on the transferable period and the transferable power range.

[0012] Secondly, embodiments of the present invention provide an electricity optimization system, comprising: The electricity price forecasting module is used to forecast electricity prices based on electricity price-related data and external electricity consumption impact data of the target area, using a pre-trained electricity price forecasting model to obtain electricity price forecast data. The electricity price-related data includes: clearing price data and electricity price data of the target area within a set time period; the external electricity consumption impact data includes: meteorological data, calendar characteristic data, and electricity supply-demand ratio of the target area within the set time period. The load characteristic construction module is used to perform cluster analysis on the electricity billing data of each user in the target area and construct the load characteristic profile of the corresponding user. The electricity dispatch module is used to perform electricity consumption simulation and extrapolation based on the electricity price forecast data and the load characteristic profile, with the goal of minimizing electricity costs, to obtain the target electricity load data for the corresponding user within a future set time, and to generate the electricity consumption optimization strategy for the corresponding user within the future set time based on the target electricity load data.

[0013] Thirdly, embodiments of the present invention provide an energy optimization device, comprising: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the energy optimization method as described in any one of the first aspects.

[0014] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the power optimization method as described in any one of the first aspects.

[0015] Compared to existing technologies, the present invention provides an electricity optimization method, system, device, and storage medium. First, based on electricity price-related data and external electricity consumption impact data for a target area, an electricity price prediction model is used to predict electricity prices, resulting in predicted electricity price data. The electricity price-related data includes clearing price data and electricity price data for the target area within a set time period. The external electricity consumption impact data includes meteorological data, calendar characteristic data, and electricity supply-demand ratio for the target area within the set time period. Then, cluster analysis is performed on the electricity settlement data of each user within the target area to construct a load characteristic profile for each user. Next, based on the predicted electricity price data and the load characteristic profile, electricity consumption simulation is performed with the goal of minimizing electricity costs, resulting in target electricity load data for the corresponding user within a set future time period. Based on the target electricity load data, an electricity optimization strategy for the corresponding user within the set future time period is generated. The present invention achieves user-side load collaborative optimization and electricity dispatch optimization by integrating multi-source characteristic electricity price prediction, user load characteristic profile clustering, and electricity consumption simulation optimization. Attached Figure Description

[0016] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of an electricity optimization method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the system architecture of the power optimization system provided in the embodiments of the present invention; Figure 3 This is a schematic diagram of the overall process of power optimization provided in the embodiments of the present invention; Figure 4 This is a structural block diagram of an electricity optimization system provided in an embodiment of the present invention; Figure 5 This is a structural block diagram of an electricity optimization device provided in an embodiment of the present invention. Detailed Implementation

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

[0019] It is understood that the various numerical designations used in the embodiments of this invention are merely for descriptive convenience and are not intended to limit the scope of this application. The order of the process numbers does not imply the order of execution; the execution order of each process should be determined by its function and internal logic.

[0020] In embodiments of the invention, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, without necessarily requiring or implying any such actual relationship or order between these entities or operations. The terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising" does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element. The term "a plurality or several" refers to two or more.

[0021] See Figure 1 , Figure 1 This is a flowchart of an electricity optimization method provided by an embodiment of the present invention. The electricity optimization method specifically includes: S11: Based on the electricity price data and external electricity consumption impact data of the target area, electricity price prediction is performed using a pre-trained electricity price prediction model to obtain electricity price prediction data; wherein, the electricity price data includes: clearing price data and electricity price data of the target area within a set time period; the external electricity consumption impact data includes: meteorological data, calendar feature data and electricity supply-demand ratio of the target area within a set time period; S12: Perform cluster analysis on the electricity billing data of each user in the target area to construct a load characteristic profile of the corresponding user; S13: Based on the electricity price forecast data and the load characteristic profile, conduct electricity consumption simulation and deduction with the goal of minimizing electricity costs to obtain the target electricity load data for the corresponding user within a future set time, and generate the electricity consumption optimization strategy for the corresponding user within the future set time based on the target electricity load data.

[0022] It should be noted that the power optimization method described in this embodiment of the invention can be executed by a power optimization system deployed on terminal devices such as servers and computers. This system is logically divided into four layers: a data access layer, a core algorithm layer, a business logic layer, and an application interaction layer, such as... Figure 2 As shown.

[0023] The data access layer is responsible for connecting with external data sources, mainly through the following methods: first, the system crawls data from the power trading center; second, users manually import data. The purpose of the data access layer is to capture or receive officially disclosed historical electricity price data, such as time-of-use pricing data, nodal marginal pricing data, clearing price data, users' historical electricity load settlement curves (i.e., electricity settlement data, also known as historical electricity consumption curves), historical external electricity consumption impact data, such as meteorological data, calendar characteristic data (such as holidays), and electricity supply-demand ratios. Its underlying logic is based on a database module, used to store cleaned historical data, user profiles, and model parameters.

[0024] For example, the system periodically retrieves historical clearing price data, historical electricity price data (i.e., historical time-of-use price data), and historical electricity load settlement curves for designated users in the target area from the power trading center server via a software interface. Furthermore, it can also retrieve meteorological data (such as weather data) and calendar feature data such as holidays from the internet.

[0025] Clearing price data refers to the historical records of market clearing prices for each trading period (day-ahead / intraday / real-time) in the electricity market, which are the marginal settlement prices when supply and demand are in balance; electricity settlement curve refers to the official curve used for electricity bill settlement, which records the actual electricity consumption / load of users at different times.

[0026] Electricity price data refers to the end-user electricity price, which can be understood as the actual settlement price of electricity consumption. It records the settlement price curve of the user's actual time-of-use electricity consumption / load.

[0027] The electricity supply-demand ratio reflects the proportional relationship between the total electricity available from the generation side (total supply) and the actual electricity demand from the user side (total demand) in the electricity market. A supply-demand ratio greater than 1 indicates ample supply (electricity prices are likely to fall), while a ratio less than 1 indicates tight supply (electricity prices are likely to rise). For example, the electricity supply-demand ratio = total supply ÷ total demand. The total supply is obtained by summing the effective electricity that all power generation companies have declared and committed to generating within the statistical forecast period; the total demand is obtained by summing the total electricity demand declared by all electricity users within the same period. It is understood that the electricity supply-demand ratio is typically calculated and updated on a rolling basis based on historical data, unit maintenance plans, load forecasts, etc.

[0028] Core Algorithm Layer: This is the computational core of the system, comprising an electricity price prediction engine, a load characteristic analysis engine, and cost simulation calculation. The electricity price prediction engine uses a time-series-based electricity price prediction model to predict future electricity prices; the load characteristic analysis engine can perform load profile analysis (such as cluster analysis) on users' historical electricity load settlement curves, identify the ratio of basic rigid load to adjustable elastic load, and construct user load characteristic profiles; cost simulation calculation.

[0029] Business logic layer: Responsible for generating specific electricity consumption optimization strategies, including a strategy optimization engine, which, based on historical data and meeting users' basic electricity consumption constraints, uses the minimization of electricity costs as the objective function to deduce the optimal electricity consumption optimization strategy; it also includes a risk assessment module, which is used to assess the potential impact of electricity price forecast deviations on cost optimization.

[0030] Application Interaction Layer: Provides a visual interface for end users, displaying electricity price forecasts, electricity budget forecasts, and optimization suggestion reports (i.e., electricity optimization strategies) through mobile terminal display modules, such as specific time periods and electricity volumes for peak shifting and valley filling, and the effectiveness of energy optimization.

[0031] This invention integrates multi-source data for electricity price prediction and user load characteristic profile clustering. Based on the predicted electricity price data and user load characteristic profiles, it performs electricity consumption simulation and optimization, enabling collaborative optimization of user-side load. This allows for the generation and visualization of electricity dispatch optimization strategies that integrate electricity price prediction and user electricity consumption behavior, providing reliable data support and decision-making basis for electricity optimization. Ultimately, it enables peak shaving and valley filling of user-side load, peak avoidance and price avoidance, effectively reducing user electricity costs, while smoothing the peak-valley difference of the power grid load and improving the efficiency of power resource allocation and the level of precision in electricity dispatch.

[0032] In one optional embodiment, electricity price prediction data is obtained by using a pre-trained electricity price prediction model based on electricity price-related data of the target area and external electricity consumption impact data, including: Data preprocessing is performed on the electricity price-related data and the external electricity consumption impact data; Correlation analysis was performed on the preprocessed electricity price-related data and external electricity consumption impact data to screen out electricity price-related features and external electricity consumption impact features whose correlation with electricity price meets the preset correlation conditions. The external electricity consumption impact features and the electricity price related features are time-aligned to construct an input feature vector; The input feature vector is fed into the electricity price prediction model to predict electricity prices, thereby obtaining electricity price prediction data.

[0033] In this embodiment of the invention, the electricity price prediction model adopts an ARIMA-LSTM hybrid model architecture. This model combines the linear fitting ability of the Autoregressive Integrated Moving Average (ARIMA) model with the nonlinear feature capture ability of the Long Short-Term Memory (LSTM) network. It also incorporates feature factors such as weather, holidays, and electricity supply-demand ratio, and combines historical clearing price data and historical electricity price data disclosed by the power trading center to achieve accurate electricity price prediction.

[0034] For example, the electricity price forecasting model adopts a cascaded hybrid architecture of "ARIMA model pre-fitting + LSTM model post-correction". First, the ARIMA model is used to capture the linear trend and seasonality of electricity price-related data. Then, the LSTM model is used to learn the nonlinear characteristics in the ARIMA model fitting residuals. The advantages of the two models are combined to output the final forecast result, as detailed below: ARIMA Model: The ARIMA(p, d, q) model is selected. Its core function is to capture the linear trend, seasonal fluctuations, and autocorrelation of historical electricity prices and historical clearing prices. Its functional expression is as follows: (1); in, This represents the historical electricity price data at time t (core input data, taken from the historical clearing prices disclosed by the power trading center). This represents a d-order difference operator used to eliminate the non-stationarity of time series corresponding to historical clearing prices. The value of d is 1 or 2, and is determined by the ADF (Augmented Dickey-Fuller) stationarity test. Represents a constant term; The numbers represent the autoregression coefficients, and Tables 1-p below represent the autoregression order, which is determined by the PACF (Partial Autocorrelation Function) plot. These represent the moving average coefficients, and 1 to q represent the moving average order, which are determined through the ACF chart. This represents the random error term at time t (i.e., the ARIMA fitting residual, which is a nonlinear feature carrier and is input into the subsequent LSTM model).

[0035] The ARIMA model has no complex network layers and only includes three core steps: differencing, autoregressive calculation, and moving average calculation. After model training and updating of model parameters (i.e., by adjusting the autoregressive coefficients and moving average coefficients in the above formula (1)), the predicted value is obtained after differencing. Performing inverse difference restoration can output the final linear prediction value. and fitting residuals .

[0036] LSTM Model: Employing a standard 3-layer LSTM network structure to capture nonlinear features in ARIMA residuals, while incorporating external features such as weather, holidays, and electricity supply-demand ratio to correct linear prediction bias. The network structure is: Input layer → LSTM hidden layer (e.g., 2 layers, with 64 and 32 neurons respectively) → Fully connected output layer (1 neuron, outputting the nonlinear correction value). Its function expression is as follows: Forgotten Gate: (2); Input Gate: , (3); Cell state: (4); Output gate: , (5); Final revised value: (6); in, This represents the input vector of the LSTM model (including the ARIMA fitting residuals). Weather characteristic values ​​at time t Holiday characteristic values ​​at time t Electricity supply and demand ratio at time t ); These represent the weight matrices of each layer; These represent the bias terms for each layer; This represents the sigmoid activation function; Represents element-wise multiplication; Indicates the output of the LSTM hidden layer; Indicates the output of the forget gate; Input gate output; Candidate cell status; Indicates the state of LSTM cells; Indicates the output of the output gate; This represents the nonlinear correction value output by the LSTM model.

[0037] Then, through fusion prediction, the linear prediction values ​​of ARIMA are... Nonlinear correction value of LSTM The weighted average is then used to obtain the final predicted electricity price, as shown below: (7); in, The fusion weights are represented, for example, values ​​of 0.6 to 0.7, and are determined through cross-validation, prioritizing the accuracy of linear trends. Let t be the final time-sharing predicted electricity price.

[0038] The input data for the entire training and inference phases of the electricity price prediction model mainly consists of the electricity price-related data and external electricity consumption impact data collected above. In order to ensure data standardization and adapt to the model input requirements, the collected data needs to be preprocessed. The data processing flow is the same for the training and inference phases, only the data range is different. For example, historical data is used in the training phase and real-time data is used in the inference phase.

[0039] The following preprocessing steps are performed on the historical electricity price data (time granularity of 15 minutes / 1 hour, consistent with the power trading center's clearing cycle) for the specified target area over the past 1-3 years, and the historical clearing prices of the spot market disclosed by the power trading center (time granularity consistent with historical electricity price data, serving as core label data for model training). Simultaneously, the daily maximum temperature, minimum temperature, and precipitation probability (from publicly available meteorological data) for the corresponding time period are collected as historical meteorological data; holiday characteristics (binary encoding: holiday = 1, weekday = 0; adjusted holiday days = 0.5) are collected as historical calendar characteristic data; and historical electricity supply-demand ratio (from supply-demand balance data disclosed by the power trading center) are also collected. Data cleaning: The collected historical electricity price data, historical clearing prices, historical meteorological data, and historical electricity supply-demand ratio are cleaned to remove outliers and / or fill in missing values ​​to ensure data integrity and accuracy. For example, the 3σ principle is used to remove extreme outliers in electricity prices and supply-demand ratios; linear interpolation is used to fill in a small number of missing data in various types of data; and duplicate values ​​in various types of data are deleted, such as clearing prices for the same period that are repeatedly disclosed by the power trading center.

[0040] Data standardization: The Z-score standardization method is used to unify the dimensions of all data. The formula is as follows: ,in, The original data, This is the historical average of this type of data. This represents the historical standard deviation of this type of data; thus, standard time series data can be formed.

[0041] It should be noted that the same set of data processing methods is used in both training and inference processes. and , and It can be calculated based on the training set during the training phase, thus avoiding data leakage.

[0042] Feature engineering: Pearson correlation analysis is used to screen the correlations of the above-mentioned data types. For example, by calculating the Pearson correlation coefficients between features in the time series data of historical electricity prices and features in the time series data of other types of data, electricity price-related features (such as electricity price features and clearing price features) and external electricity consumption impact features (such as meteorological features, calendar features, and electricity supply-demand ratio features) that meet the correlation criteria are screened. Features with low correlation to electricity prices are eliminated, for example, the correlation criterion is the absolute value of the correlation coefficient. If the correlation coefficient is 0.3, features with an absolute value less than 0.3 are removed, and the core features are retained. Then, the time series data of the filtered external electricity consumption impact data and the time series data of electricity price related data are aligned on the time axis to construct a unified input feature vector.

[0043] Data partitioning (training phase only): The standardized data processed above is divided into training set, validation set, and test set according to a preset ratio (e.g., 7:2:1). The training set is used for model parameter training (ARIMA order determination, LSTM weight update); the validation set is used to adjust model hyperparameters (e.g., p, d, q of ARIMA, number of neurons in LSTM, fusion weight α); the test set is used to verify the final prediction accuracy of the model (using root mean square error RMSE as the evaluation metric).

[0044] Data input adaptation for model training / inference: During the training phase, the ARIMA model is input with historical electricity price data and historical clearing price data (standardized time series data) from the training set. It is trained to obtain a linear fitting model and outputs linear prediction values ​​and residuals. The LSTM model is input with the ARIMA fitting residuals (standardized) from the training set plus external electricity consumption impact data (standardized time series data). The actual values ​​of the residuals are used as labels to train and correct the model. The fusion weights α are adjusted through the validation set to minimize the RMSE between the fused electricity price prediction value and the corresponding historical electricity price, thus completing the model training and obtaining the trained electricity price prediction model.

[0045] During the inference phase, the system inputs real-time collected electricity price data (standardized time series data) for the previous n time periods (e.g., the week before the current time), clearing prices for the previous n time periods (standardized time series data), and external electricity consumption impact data for the previous n time periods and a future set time (standardized time series data). An input feature vector is constructed according to the feature fusion rules of the training phase. First, a linear prediction value is output through the ARIMA model, then a nonlinear correction value is output through the LSTM model. After weighted fusion, the electricity price prediction data for the future preset time period (e.g., the week before the current time) is obtained, such as the time-of-use electricity price prediction curve for the future preset time period. Simultaneously, the probability density distribution of the prediction results is output, providing data support for risk assessment.

[0046] Among them, the electricity price prediction model uses the root mean square error (RMSE) as the loss function for model training to ensure prediction accuracy; N is the sample size. For predicted values, The prediction results include not only the predicted point values ​​but also the probability density distribution, which is used to assess high-price risks at the business logic layer and improve the reliability of the strategy.

[0047] The embodiments of this invention employ an ARIMA-LSTM hybrid model that can fully learn the temporal patterns of electricity prices themselves and the correlation effects of external electricity consumption factors, resulting in a more accurate fit to the actual market clearing price change trend and significantly improving the accuracy, reliability, and generalization ability of electricity price prediction results.

[0048] In one optional embodiment, cluster analysis is performed on the electricity billing data of each user within the target area to construct a load characteristic profile for the corresponding user, including: For each user's electricity billing data, the K-means clustering algorithm is used to extract the daily load curves for different daily scenarios from the electricity billing data; The daily load curve is decomposed to determine the rigid load and elastic load for each time period under the corresponding daily type scenario; Based on the rigid and flexible loads at different times under different day types, a load characteristic profile of the corresponding user under different day types is constructed.

[0049] Furthermore, the daily load curve is decomposed to determine the rigid and flexible loads for each time period under the corresponding daily scenario, including: For the daily load curves under different day types, a moving average is performed on the daily load curves to obtain the load baseline; The load baseline is filtered by quantiles to obtain the rigid load for each time period under the corresponding day type scenario; Based on the daily load curve and the rigid load of each time period, calculate the elastic load of each time period under the corresponding daily type scenario.

[0050] The calculation process for the elastic load transferable time window includes: Based on the time period of the rigid load, the time period of the elastic load, and the preset power consumption constraints, calculate the transferable time period of the elastic load in each time period; Based on the rigid and flexible loads for each time period, and the actual load for the corresponding time period in the daily load curve, calculate the transferable power range for each time period. The transferable time window for flexible loads is determined based on the transferable period and the transferable power range.

[0051] In this embodiment of the invention, by statistically analyzing users' electricity billing data, such as historical electricity consumption curves, users' electricity consumption habits are identified, and the electricity load is decomposed into rigid load and flexible load to calculate the transferable time window of the flexible load, providing data support for subsequent electricity consumption optimization.

[0052] Furthermore, this embodiment of the invention achieves load identification through a virtual decomposition method, eliminating the need for on-site sensor installation. For example, based on the daily load curve in a user's electricity settlement data (i.e., historical electricity consumption curve) provided by the power trading center, the system uses a clustering algorithm to divide historical days into different day types such as typical workdays, typical rest days, and peak days. Under each day type scenario, a rigid load benchmark is established, and the portion of the daily load curve that fluctuates above the rigid load benchmark is considered as the flexible load space. When generating strategies, the system does not vaguely suggest "saving electricity," but rather calculates specific transferable electricity values ​​based on predicted low-price periods, forming quantifiable and implementable electricity cost optimization suggestions adapted to different users' electricity consumption scenarios. The specific load clustering analysis process is as follows: Pre-setting daily load characteristics and clustering conditions for different day types: This embodiment of the invention uses the K-means clustering algorithm, combined with the time distribution characteristics of the user's historical electricity consumption curve, load peak / valley values, load fluctuation amplitude and other indicators, to divide historical days into 3 typical scenarios. The daily load characteristics and clustering conditions for each scenario are as follows, to ensure that the scenario division conforms to the user's actual electricity consumption patterns.

[0053] Typical weekday scenario: Time determination: Non-holidays and non-adjusted workdays (determined by combining holiday feature codes, with holidays = 0 and adjusted workdays = 0.5). Daily load characteristics: The daily load exhibits a "double peak and double trough" distribution. The peak load during the morning peak (7:00-9:00) is 1.3 to 1.5 times the daily average load, the peak load during the evening peak (17:00-19:00) is 1.4 to 1.6 times the daily average load, and the load during the midday trough (12:00-14:00) and night trough (00:00-06:00) is 0.6 to 0.8 times the daily average load. Clustering threshold (clustering conditions): Intraday load fluctuation coefficient ≤ 0.35, and peak load ratio ≥ 18%.

[0054] Among them, the daily average load = total load of all time periods during the day / sum of all time periods during the day; the daily load fluctuation coefficient = daily load standard deviation / daily average load; the peak load percentage = sum of daily peak loads / daily total load, and the sum of daily peak loads = morning peak load + evening peak load.

[0055] Typical weekend scenario: Time determination: public holidays (public holidays = 1) or adjusted workdays (adjusted workdays = 0.5); Daily load characteristics: There is no obvious double peak in the daily load, the load curve is flat, the peak occurs between 10:00 and 16:00, and the peak is 1.1 to 1.2 times the daily average load. The night load (22:00 to 06:00) is consistently at a low level, which is 0.5 to 0.7 times the daily average load. Clustering threshold (clustering conditions): Intraday load fluctuation coefficient ≤ 0.25, and no obvious early morning peak or late evening peak, i.e., peak load ratio < 15%.

[0056] Peak day scenario: Time determination: Regardless of weekdays / restdays, it takes into account weather characteristics or special industrial production periods, such as days with a maximum daily temperature of ≥35℃ or ≤5℃, or days during the factory's monthly peak production period; Daily load characteristics: The overall daily load is relatively high, with the peak load being 1.6 to 1.8 times the daily average load and the peak duration being ≥3 hours. The trough load is 0.7 to 0.9 times the daily average load, with no obvious trough period. Clustering threshold (clustering conditions): Daily average load ≥ 1.2 times the historical daily average load, and daily load fluctuation coefficient ≥ 0.4.

[0057] Based on the daily load characteristics and clustering conditions under different daily scenarios set above, the historical electricity consumption curves of each user are clustered to extract the daily load curves under different daily scenarios. For example, firstly, the daily load curves of the user over the past 12 months are extracted, and the features such as the intraday average load, intraday load fluctuation coefficient, peak load and peak load percentage, valley load and peak load percentage of each daily load curve are calculated; then, the calculated features are clustered using the K-means algorithm (K=3), and the cluster centers are iteratively optimized until the clustering error (such as the silhouette coefficient) is ≥0.8, thus completing the division of different daily scenarios and obtaining the daily load curves under different daily scenarios.

[0058] Then, load decomposition is performed on the daily load curves under different day types to construct a load characteristic profile of the user under different day types.

[0059] Taking a small to medium-sized manufacturing user (typical weekday scenario) as an example, the user's historical electricity consumption curve is the daily load curve of the past 12 months (time granularity 1 hour, a total of 24 data points / day). The specific process of constructing its load characteristic profile is as follows: Data preprocessing: Mark the scenario corresponding to each daily load curve (typical workday / rest day / peak day, divided according to the clustering conditions above) to extract the daily load curve of the user under different scenarios, remove outliers (such as load drop caused by equipment failure), and perform normalization processing; Load decomposition identifies rigid and elastic loads for different scenarios. The specific process is as follows: Historical average load and standard deviation are calculated for each scenario to determine the daily load curve and calculate the load baseline for different scenarios. For each scenario s and each time period t, the average historical load over M days is calculated. and standard deviation Abnormal data (such as samples with a load of 0 or sudden load fluctuations exceeding 30%) are removed to ensure data reliability.

[0060] The load baseline is calculated by moving average the daily load curves for all typical workdays to determine the average load at each time point within the day. Based on the average load of each time period during the day This allows the load baseline to be established. This represents the actual load value (unit: kW) on day j and time period t under scenario s. Indicate the scenario type (1=typical weekday, 2=typical weekend, 3=peak day); M represents the intraday time period (0-23 hours, with each hour being a time period); M represents the number of days corresponding to scenario s (i.e., the number of daily load curves under scenario s, M≥30, to ensure sufficient sample size and improve baseline accuracy).

[0061] Based on the load baseline, the rigid load for each time period (also known as the rigid load baseline) is determined. The rigid load baseline refers to the minimum electricity load (non-adjustable) that a user must be guaranteed under a specific scenario. It is calculated using a scenario-specific approach combined with moving averages and quantile correction. The calculated load baseline (i.e., average load) is corrected for the standard deviation to obtain the corrected load set; the 85th quantile of this set is the rigid load baseline for scenario s and time period t. The calculation of the rigid load baseline is as follows: (8); in, Represents the rigid load baseline value (unit: kW) for scenario s and time period t; This represents the average load for scenario s and time period t; This represents the standard deviation of the load for scenario s and time period t (used to correct the average load and avoid the influence of outliers). () represents the quantile function; 0.85 represents taking the 85th quantile of the corrected load value to ensure that the rigid load baseline value covers more than 90% of the minimum load demand and avoid misjudging the flexible load as a rigid load.

[0062] By combining the actual load of each time period with the rigid load of the corresponding scenario from the daily load curves of the same scenario over the previous n time periods (e.g., the week before the current time), the elastic load is obtained. For example, for each scenario s and each time period t, the actual load is used... Subtract the rigid load baseline for the corresponding period The elastic load is obtained; if the difference is negative, it is set to 0 to ensure that the elastic load is ≥ 0. That is, the elastic load... Among them, the elastic load is ≥0. If the difference is negative, it means that the actual load is lower than the baseline. In this case, the elastic load of the corresponding period is directly set to 0, indicating that there is no elastic adjustable load.

[0063] For example, taking the 12:00-13:00 period (t=12) of a typical working day (s=1) for the aforementioned manufacturing user, with a typical working day of 60 days, the average load is 720kW, the standard deviation is 30kW, and the corrected load set is [690, 750]. Taking the 85th quantile, it is 735kW, i.e., P baseline (1, 12) = 735kW; the actual load on a certain day during that period is 900kW, then the elastic load is 900 - 735 = 165kW.

[0064] Calculate the flexible load transferable time window. The flexible load transferable time window refers to the time range and adjustment range within which flexible load can be transferred from the current time period to other time periods. It consists of two core indicators: the transferable time period and the transferable electricity amount.

[0065] First, determine the non-transferable period. For example, by combining user electricity consumption scenarios (such as production processes), rigid load periods (such as the core production period of manufacturing from 7:00 to 19:00) can be marked. Loads that cannot be transferred during these periods are called non-transferable periods. ; Then, calculate the transferable time period. Taking a single elastic load's time period t (e.g., 12:00-13:00) as an example, its transferable time period is calculated as follows: (9); in, The transferable time period (also known as the set of transferable time periods or the range of transferable time periods, in hours, 0-23 hours within a day) represents the flexible load corresponding to time period t. express One of the transferable time periods; L represents the maximum number of hours that can be transferred forward (L≤6, to avoid cross-day transfer and adapt to the user's intraday electricity consumption constraints); R represents the maximum number of hours that can be transferred backward (R≤6, to avoid cross-day transfer). Rigid load non-transferable periods (also known as non-transferable period sets, such as core production periods 7:00-9:00 and 17:00-19:00, during which loads cannot be transferred).

[0066] Understandably, the values ​​of L and R are set according to the user's power consumption constraints, and the upper limit of L and R is usually 6.

[0067] Next, the range of transferable electricity is calculated. This refers to the maximum amount of flexible load that can be transferred in each time period t; the specific calculation is as follows: (10); in, This represents the actual flexible load electricity during time period t. ; Indicates the transferable period The rigid load baseline; Indicates the transferable period The actual load; This represents the sum of the flexible loads for all transferable periods (i.e., the maximum transferable amount of electricity).

[0068] By integrating the transferable time period and the transferable electricity amount, a flexible load transferable time window can ultimately be formed for that period.

[0069] For example, taking the 12:00-13:00 time period (t=12) of a typical working day for the aforementioned manufacturing user as an example, L=3, R=4, T rigid ={12-3, 12+4}, then (12)=={9, 10, 11, 13, 14, 15, 16}; P during this period elastic (12)==200kW, the sum of the elastic loads in all transferable periods is 250kW, then P transfer (12) = min(200, 250) = 200kW, meaning that the 200kW flexible load during this period can be transferred to any time period between 9:00-11:00 and 13:00-16:00.

[0070] Based on the distribution of rigid and flexible loads in different user scenarios, a load characteristic profile of the user in different scenarios is constructed. For example, the load characteristic profile of this user on a typical workday is: "Rigid load accounts for 75%, and flexible load accounts for 25%; rigid load is mainly concentrated between 7:00-19:00 (core production period), and flexible load is mainly concentrated between 12:00-14:00 (lunch break period) and 20:00-23:00 (nighttime auxiliary production period)."

[0071] This invention uses K-means clustering to divide scenarios and decomposes rigid and elastic loads according to scenarios, objectively depicting the multi-dimensional electricity consumption behavior patterns of users. It can accurately quantify basic electricity consumption and adjustable load potential, construct a standardized load characteristic profile, and provide a reliable basis for electricity price forecasting, demand-side management, and electricity consumption characteristic mining.

[0072] In one optional embodiment, based on the electricity price forecast data and the load characteristic profile, an electricity consumption simulation is performed with the goal of minimizing electricity costs to obtain the target electricity load data for the corresponding user within a set future time period, including: The first time interval corresponding to the predicted electricity price that is higher than the set first electricity price threshold is selected from the electricity price prediction data; Based on the load characteristic profile, calculate the transferred load amount for the first time interval; In a pre-built virtual settlement environment, the transferred load is transferred from the first time interval to the corresponding transferable time period, and a simulated settlement is performed according to the electricity price forecast data to obtain the final transferred load with the goal of minimizing electricity costs. Based on the final transferred load and the daily load curve, the target electricity billing data for the corresponding users within a set future time period is determined.

[0073] Specifically, based on the load characteristic profile, the amount of transferred load in the first time interval is calculated, including: Extract the elastic load corresponding to the first time interval from the load characteristic profile, and use it as the upper limit of the transferable load for the first time interval; Based on the load characteristic profile, calculate the total load capacity of the elastic load transfer time window in the first time interval; The minimum value between the upper limit of transferable load and the total load capacity is selected as the amount of transferable load for the first time interval.

[0074] This invention constructs a virtual settlement environment within the system software, establishes a logical relationship between price and electricity consumption through simulation, and ultimately generates an electricity consumption optimization strategy.

[0075] Understandably, electricity price forecast data is time-series data recording predicted electricity prices for different time periods. When it is predicted that the period from 10:00 to 12:00 tomorrow will be a high-price period (i.e., the first time interval, such as when the predicted electricity price is higher than the set second electricity price threshold), the system will automatically call the cost simulation calculation engine to simulate how much the total electricity bill will decrease if the user shifts the flexible load of this period to 23:00-01:00. The system directly presents this calculation result to the user, so that the price signal is directly transformed into electricity consumption behavior guidance, load control instructions, and executable optimization strategies. This software mapping relationship between price, cost, and strategy achieves effective transmission of price signals to electricity consumption behavior without relying on hardware control.

[0076] The transferred load during the high-price period (i.e., the first time interval) refers to the transferable flexible load during that period. It must simultaneously meet two core constraints: it must not exceed the total flexible load, nor exceed the maximum transferable electricity that the flexible load transfer window can accommodate. In other words, the transferred load during the high-price period ≤ the flexible load during that period (i.e., the upper limit of transferable load) ≤ the maximum transferable electricity that the flexible load transfer window can accommodate (i.e., the total load capacity). The specific method for determining the transferred load during the high-price period (i.e., the first time interval) is as follows: Determine the high-price periods and their corresponding flexible load. First, predict electricity price data for a set time period using an electricity price forecasting model, and mark the high-price periods (i.e., the first time interval, such as 10:00-12:00 tomorrow) where the predicted electricity price is higher than the set price threshold. Then, use the rigid load baseline calculated above for the same day type scenario to determine the high-price periods. Calculate the elastic load during this high-price period. This is the upper limit of the elastic load that can be transferred during this high-price period.

[0077] Determine the maximum transferable electricity volume that can be accommodated within the flexible load transfer window. For this high-price period, calculate the set of transferable periods based on the flexible load transfer window calculation method described above. (i.e., the period during which the transferred load can be accommodated in the low-price area); then through The sum of the remaining capacity for all transferable periods is the maximum transferable electricity that can be accepted within the flexible load transferable time window. This prevents the load during the transferable period from exceeding its rigid baseline after transfer, thus affecting the user's normal energy consumption.

[0078] Determine the final load transfer volume. The final transferable load volume during the high-price period is selected from the flexible load volume calculated above for that high-price period. The minimum value among the maximum transferable electricity volume that can be accepted within the flexible load transfer time window is used, and fine-tuned in conjunction with minimizing electricity costs (i.e., maximizing electricity cost savings). The specific fine-tuning formula is as follows: (11); in, This represents the final transferred load during the high-price period t (unit: kW). This represents the total elastic load during the high-price period t (Constraint 1: It cannot exceed the upper limit of the elastic load). This indicates the maximum amount of electricity that can be transferred within the flexible load transfer time window (Constraint 2: Cannot exceed the upper limit of acceptance). This represents the optimal transfer amount corresponding to maximizing electricity cost savings (calculated by a cost simulation engine within two constraints to ensure maximum reduction in electricity costs after the transfer).

[0079] Based on the aforementioned predicted electricity price data and load characteristic profile, an objective function can be constructed with the goal of minimizing cost: Min(Total Electricity Cost) = Σ(Predicted Electricity Price × Optimized Electricity Load). The constraints are: total electricity consumption remains constant, rigid load conditions must be met, and the load must be within the limits of the transferable periods and transferable electricity volume of the flexible load. Solving this function using linear programming or heuristic algorithms yields the theoretically optimal electricity load curve (i.e., the target electricity billing data). This optimal load curve records the optimized electricity load for each period within a set future timeframe. Using the optimal load curve and the aforementioned objective function, the expected electricity cost under this optimal load curve can be calculated. Comparing this to the expected electricity cost under the original daily load curve before load transfer (Σ(Predicted Electricity Price × Unoptimized Electricity Load)), the expected reduction in electricity cost under the optimal load curve can be obtained.

[0080] Furthermore, the load transfer during high-price periods will be prioritized from low-price periods (when the predicted electricity price is below the second electricity price threshold) within the set of transferable periods. For example, if the transferable time windows are 9:00-11:00 and 13:00-16:00, and 13:00-15:00 is predicted to be a high-price period and 9:00-11:00 a low-price period, then the flexible load from 13:00-15:00 will be prioritized for transfer to 9:00-11:00.

[0081] Taking a manufacturing user with 150kW of flexible load during the 10:00-11:00 (high-price period) as an example, which can be transferred to the 23:00-01:00 (low-price period), with a final transfer of 80kW (subject to the capacity limit of the transferable period), the high-price period flexible load transfer strategy is as follows: Scenario-based flexible load shifting strategy (adapting to different electricity consumption scenarios): Based on three categories of user electricity consumption scenarios divided by K-means clustering, a targeted peak shifting and valley filling strategy is formulated to match the actual electricity consumption patterns of users, as detailed below: Typical weekday scenario: In response to the "double peak and double valley" load characteristics, the flexible load (such as auxiliary production equipment) of the morning peak (7:00-9:00) and evening peak (17:00-19:00) is transferred to the low-price period of the afternoon valley (12:00-14:00) and night valley (00:00-06:00); Typical rest day scenario: The load curve is flat, and the flexible load (such as air conditioning and lighting) from 10:00-16:00 (peak period, if it is high price) is transferred to the low price period from 22:00-06:00; Peak day scenario: Overall load is relatively high. Prioritize transferring non-core flexible loads (such as logistics electricity) to the low-price period at night to avoid load overlap during peak hours and reduce electricity costs.

[0082] Dynamic peak-shaving strategy based on load baseline (ensuring energy security): Taking the rigid load baseline as the core constraint, the strategy dynamically adjusts the pace of flexible load transfer, balancing peak-shaving and valley-filling effects with user energy needs, as detailed below: Rigid load guarantee: Prioritize ensuring sufficient supply of rigid load during core periods (such as manufacturing 7:00-19:00), and only transfer flexible load during this period (the remaining load after transfer shall not be lower than the rigid load baseline). Dynamic adjustment: If the load during the predicted high-price period is close to the rigid load baseline (without flexibility), then the peak shifting operation for that period will be suspended; if there is sufficient transferable capacity during the low-price period, the amount of load transferred can be appropriately increased to maximize the valley filling effect. By shifting flexible loads from high-price periods to low-price periods, the optimal global electricity consumption arrangement can be obtained under compliance constraints. This enables peak shaving and valley filling on the load side, smoothing grid fluctuations, and maximizing the avoidance of peak periods with high electricity prices, effectively reducing users' electricity costs.

[0083] Optimization suggestion generation and feedback. The system compares the optimal electricity load curve with the user's original electricity usage habits (i.e., the original daily load curve) and generates a difference report. The system pushes specific electricity optimization strategies to the user, such as suggesting reducing the load by 500kW during the high-price period of 14:00-16:00 and shifting it to the low-price period of 02:00-04:00, and displaying the expected amount of electricity cost savings.

[0084] Furthermore, the system can record user adoption status and iterate the model by comparing subsequent settlement data and combining data updates. The system's feedback mechanism relies on the disclosure of new data from the power trading center. The system is set up with a scheduled task to obtain the actual settlement electricity price and actual electricity consumption curve of the previous day. The actual value is compared with the predicted electricity price for the same day predicted by the above-mentioned electricity price prediction model and the expected electricity charge obtained from simulation. The prediction deviation and electricity charge reduction deviation are calculated. These deviation data are used as new training samples to input into the core algorithm layer, automatically updating the electricity price prediction model and load characteristic profile parameters, thereby continuously improving the accuracy of the recommendations over time.

[0085] Compared with the prior art, the beneficial effects of the embodiments of the present invention are as follows: (1) The embodiments of the present invention adopt a pure software architecture, which does not require the installation of any hardware sensors or actuators on site. It can directly access the publicly disclosed data and user historical settlement data of the power trading center for power consumption optimization analysis. Users do not need to change their existing power facilities, but only need to access the software account to use it, which greatly reduces the threshold and cost for users to participate in power market optimization.

[0086] (2) The embodiments of the present invention construct an electricity price forecasting system. Load characteristic analysis Cost simulation optimization It is recommended that the internal logical loop of the generated report be closed, such as Figure 3 As shown, while achieving electricity price forecasting, the software simulation also calculates the specific differences in electricity costs under different electricity consumption behaviors, transforming the abstract electricity price signal into cost savings that are visible to users. This effectively guides users to adjust their electricity consumption behavior, truly realizing the effective transmission of price signals to the user side, solving the problem of the disconnect between signals and electricity consumption behavior, and achieving a soft closed loop from forecasting to decision-making.

[0087] (3) Load elasticity identification based on historical data profile: Based on historical electricity consumption curves, virtual load decomposition is performed to identify rigid load baseline and elastic load space. Users’ load adjustment potential can be inferred without real-time sensing data. It is highly adaptable, which enables the system to provide reliable reference suggestions based on historical patterns even in the event of network fluctuations or data delays.

[0088] (4) Using the actual settlement data disclosed daily / monthly by the power trading center as the true value, the system automatically compares the predicted electricity price and the simulated electricity fee value to obtain the difference data. Based on these difference data, the electricity price prediction model and load characteristic parameters are automatically updated. As the usage time increases, the accuracy of the grasp of users' electricity consumption habits and regional electricity price patterns can be improved, forming a self-evolutionary ability that becomes more accurate with use, and further improving the prediction accuracy.

[0089] See Figure 4 , Figure 4 This invention provides a structural block diagram of an electricity optimization system, which includes: The electricity price prediction module 11 is used to predict electricity prices based on electricity price-related data and external electricity consumption impact data of the target area, using a pre-trained electricity price prediction model to obtain electricity price prediction data. The electricity price-related data includes: clearing price data and electricity price data of the target area within a set time period; the external electricity consumption impact data includes: meteorological data, calendar feature data and electricity supply-demand ratio of the target area within a set time period. The load characteristic construction module 12 is used to perform cluster analysis on the electricity billing data of each user in the target area and construct a load characteristic profile of the corresponding user. The power dispatch module 13 is used to perform power consumption simulation and deduction based on the electricity price forecast data and the load characteristic profile, with the goal of minimizing power consumption cost, to obtain the target power load data of the corresponding user within a future set time, and to generate the power consumption optimization strategy of the corresponding user within the future set time based on the target power load data.

[0090] In one optional embodiment, the electricity price forecasting module 11 includes: The data preprocessing unit is used to preprocess the electricity price-related data and the external electricity consumption impact data; The correlation analysis unit is used to perform correlation analysis on the preprocessed electricity price-related data and external electricity consumption impact data, and to screen the electricity price-related features and external electricity consumption impact features that meet the preset correlation conditions with the electricity price. A time alignment unit is used to align the external electricity consumption impact features with the electricity price-related features over time to construct an input feature vector. The electricity price prediction unit is used to feed the input feature vector into the electricity price prediction model to predict electricity prices and obtain electricity price prediction data.

[0091] In an optional embodiment, the load characteristic construction module 12 includes: Clustering units are used to extract daily load curves for different daily scenarios from the electricity billing data of each user using the K-means clustering algorithm. The load decomposition unit is used to decompose the daily load curve to determine the rigid load and elastic load for each time period under the corresponding daily type scenario. The load characteristic profile building unit is used to build load characteristic profiles for corresponding users in different day types based on the rigid load and elastic load of each time period under different day types.

[0092] In one optional embodiment, the load decomposition unit includes: The moving average sub-unit is used to perform moving average processing on the daily load curves under different day type scenarios to obtain the load baseline; The quantile filtering subunit is used to perform quantile filtering on the load baseline to obtain the rigid load for each time period under the corresponding day type scenario; The flexible load calculation subunit is used to calculate the flexible load for each time period under the corresponding daily type scenario based on the daily load curve and the rigid load for each time period.

[0093] In one optional embodiment, the power dispatching module 13 includes: The time interval determination unit is used to filter out the first time interval corresponding to the predicted electricity price that is higher than the set first electricity price threshold from the electricity price prediction data; The load transfer calculation unit is used to calculate the load transfer amount for the first time interval based on the load characteristic profile. The simulation settlement unit is used to transfer the load from the first time interval to the corresponding transferable time period in a pre-built virtual settlement environment, and to perform simulation settlement according to the electricity price forecast data to obtain the final load transfer with the goal of minimizing electricity costs. The electricity consumption determination unit is used to determine the target electricity consumption settlement data of the corresponding users within a future set time period based on the final transferred load and the daily load curve.

[0094] In one optional embodiment, the load transfer calculation unit includes: The transferable load upper limit determination subunit is used to extract the elastic load corresponding to the first time interval from the load characteristic profile, as the transferable load upper limit of the first time interval; The total load acceptance capacity determination subunit is used to calculate the total load acceptance capacity of the flexible load transfer time window in the first time interval based on the load characteristic profile. The load transfer determination subunit is used to select the minimum value from the upper limit of transferable load and the total load acceptance capacity as the load transfer amount for the first time interval.

[0095] In one alternative embodiment, the system includes: The transferable time period calculation module is used to calculate the transferable time period of the elastic load in each time period based on the time period in which the rigid load is located, the time period in which the elastic load is located, and the preset power consumption constraints. The transferable power amplitude calculation module is used to calculate the transferable power amplitude for each time period based on the rigid load and flexible load for each time period, as well as the actual load for the corresponding time period in the daily load curve. The flexible load transferable time window determination module is used to determine the flexible load transferable time window based on the transferable period and the transferable power range.

[0096] It should be noted that the working process of each module in the power optimization system described in the embodiments of the present invention can refer to the working process of the power optimization method described in the above embodiments, and the technical effect achieved is the same as that of the power optimization method described in the above embodiments, so it will not be repeated here.

[0097] See Figure 5 , Figure 5 This is a structural block diagram of the power optimization device provided in an embodiment of the present invention. The power optimization device includes a processor 21, a memory 22, and a computer program stored in the memory 22 and executable on the processor 21. When the processor 21 executes the computer program, it implements the steps in the various power optimization method embodiments described above, such as steps S11 to S13.

[0098] For example, the computer program may be divided into one or more modules or units, which are stored in the memory 22 and executed by the processor 21 to complete the present invention. The one or more modules or units may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the power optimization device.

[0099] The power optimization device may include, but is not limited to, a processor 21 and a memory 22. Those skilled in the art will understand that the schematic diagram is merely an example of the power optimization device and does not constitute a limitation on it. It may include more or fewer components than illustrated, or combine certain components, or different components. For example, the power optimization device may also include input / output devices, network access devices, buses, etc.

[0100] The processor 21 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor 21 is the control center of the power optimization equipment, connecting all parts of the equipment via various interfaces and lines.

[0101] The memory 22 can be used to store the computer programs and / or modules. The processor 21 implements various functions of the power optimization device by running or executing the computer programs and / or modules stored in the memory 22 and calling the data stored in the memory 22. The memory 22 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory 22 may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0102] Wherein, if the modules or units integrated into the power optimization equipment are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor 21, it can implement the steps of the various method embodiments described above. Wherein, the computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0103] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0104] The above description represents the preferred embodiments of the present invention. It should be noted that, for those skilled in the art, various improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for optimizing electricity consumption, characterized in that, include: Based on electricity price data and external electricity consumption impact data for the target area, a pre-trained electricity price prediction model is used to predict electricity prices, resulting in predicted electricity price data. The electricity price data includes: clearing price data and electricity price data for the target area within a specified time period; the external electricity consumption impact data includes: meteorological data, calendar characteristic data, and electricity supply-demand ratio for the target area within the specified time period. Cluster analysis is performed on the electricity billing data of each user in the target area to construct a load characteristic profile of the corresponding user; Based on the electricity price forecast data and the load characteristic profile, electricity consumption simulation is performed with the goal of minimizing electricity costs to obtain the target electricity load data for the corresponding user within a future set time. Based on the target electricity load data, an electricity consumption optimization strategy for the corresponding user within the future set time is generated.

2. The power consumption optimization method as described in claim 1, characterized in that, Based on electricity price data and external electricity consumption impact data for the target area, a pre-trained electricity price prediction model is used to predict electricity prices, resulting in predicted electricity price data, including: Data preprocessing is performed on the electricity price-related data and the external electricity consumption impact data; Correlation analysis was performed on the preprocessed electricity price-related data and external electricity consumption impact data to screen out electricity price-related features and external electricity consumption impact features whose correlation with electricity price meets the preset correlation conditions. The external electricity consumption impact features and the electricity price related features are time-aligned to construct an input feature vector; The input feature vector is fed into the electricity price prediction model to predict electricity prices, thereby obtaining electricity price prediction data.

3. The power consumption optimization method as described in claim 1, characterized in that, Cluster analysis is performed on the electricity billing data of each user within the target area to construct a load characteristic profile for the corresponding user, including: For each user's electricity billing data, the K-means clustering algorithm is used to extract the daily load curves for different daily scenarios from the electricity billing data; The daily load curve is decomposed to determine the rigid load and elastic load for each time period under the corresponding daily type scenario; Based on the rigid and flexible loads at different times under different day types, a load characteristic profile of the corresponding user under different day types is constructed.

4. The power consumption optimization method as described in claim 3, characterized in that, The daily load curve is decomposed to determine the rigid and flexible loads for each time period under the corresponding daily scenario, including: For the daily load curves under different day types, a moving average is performed on the daily load curves to obtain the load baseline; The load baseline is filtered by quantiles to obtain the rigid load for each time period under the corresponding day type scenario; Based on the daily load curve and the rigid load of each time period, calculate the elastic load of each time period under the corresponding daily type scenario.

5. The power consumption optimization method as described in claim 4, characterized in that, Based on the electricity price forecast data and the load characteristic profile, electricity consumption simulation is performed with the goal of minimizing electricity costs to obtain the target electricity load data for the corresponding users within a set future time period, including: The first time interval corresponding to the predicted electricity price that is higher than the set first electricity price threshold is selected from the electricity price prediction data; Based on the load characteristic profile, calculate the transferred load amount for the first time interval; In a pre-built virtual settlement environment, the transferred load is transferred from the first time interval to the corresponding transferable time period, and a simulated settlement is performed according to the electricity price forecast data to obtain the final transferred load with the goal of minimizing electricity costs. Based on the final transferred load and the daily load curve, the target electricity billing data for the corresponding users within a set future time period is determined.

6. The power consumption optimization method as described in claim 5, characterized in that, Based on the load characteristic profile, the amount of transferred load in the first time interval is calculated, including: Extract the elastic load corresponding to the first time interval from the load characteristic profile, and use it as the upper limit of the transferable load for the first time interval; Based on the load characteristic profile, calculate the total load capacity of the elastic load transfer time window in the first time interval; The minimum value between the upper limit of transferable load and the total load capacity is selected as the amount of transferable load for the first time interval.

7. The power consumption optimization method as described in claim 6, characterized in that, The calculation process for the elastic load transferable time window includes: Based on the time period of the rigid load, the time period of the elastic load, and the preset power consumption constraints, calculate the transferable time period of the elastic load in each time period; Based on the rigid and flexible loads for each time period, and the actual load for the corresponding time period in the daily load curve, calculate the transferable power range for each time period. The transferable time window for flexible loads is determined based on the transferable period and the transferable power range.

8. A power consumption optimization system, characterized in that, include: The electricity price forecasting module is used to forecast electricity prices based on electricity price-related data and external electricity consumption impact data of the target area, using a pre-trained electricity price forecasting model to obtain electricity price forecast data. The electricity price-related data includes: clearing price data and electricity price data of the target area within a set time period; the external electricity consumption impact data includes: meteorological data, calendar characteristic data, and electricity supply-demand ratio of the target area within the set time period. The load characteristic construction module is used to perform cluster analysis on the electricity billing data of each user in the target area and construct the load characteristic profile of the corresponding user. The electricity dispatch module is used to perform electricity consumption simulation and extrapolation based on the electricity price forecast data and the load characteristic profile, with the goal of minimizing electricity costs, to obtain the target electricity load data for the corresponding user within a future set time, and to generate the electricity consumption optimization strategy for the corresponding user within the future set time based on the target electricity load data.

9. A power optimization device, characterized in that, include: A processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the power optimization method as described in any one of claims 1 to 7.

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

Images