Day-ahead-real-time cooperative scheduling system of electric energy replacement load and energy storage system

Abstract

The invention provides a day-ahead-real-time cooperative scheduling system for an electric energy replacement load and an energy storage system, and relates to the field of power management. Respectively inputting the load prediction model and the power generation prediction model to obtain a day-ahead load prediction result and a day-ahead power generation prediction result, generating a day-ahead scheduling plan based on the obtained prediction results and a preset optimization target, and then obtaining actual electric energy replacement load and renewable energy power generation data; deviation information of actual data and predicted data is calculated according to a day-ahead scheduling plan, and finally the charging and discharging power of the energy storage system is adjusted according to the deviation information. Therefore, the coordinated operation efficiency of the electric energy replacement equipment cluster and the renewable energy sources can be effectively improved, and the economic operation, safety and stability of the system are effectively ensured.

Description

Technical Field

[0001] This application relates to the field of power management, and more specifically, to a day-ahead and real-time collaborative scheduling system for electric energy substitution loads and energy storage systems. Background Art

[0002] With the advancement of the "dual carbon" goal, the large-scale access of electric energy substitution loads (such as electric vehicles, hydrogen production by electricity, industrial electric heating, etc.) and the increase in the penetration rate of renewable energy, the power system faces double challenges of uncertainty on both the supply and demand sides. Electric energy substitution loads have significant time-varying and demand response characteristics, and their electricity consumption behaviors are affected by the interweaving of multiple factors such as weather, user habits, and electricity price policies; while the output of renewable energy such as wind power and photovoltaic power is strongly correlated with meteorological conditions, showing intermittency and volatility. To cope with such uncertainties, the power system needs to achieve precise matching of source-storage-load resources through the collaborative optimization of day-ahead scheduling plans and real-time dynamic adjustments.

[0003] However, in the traditional scheduling mode, the accuracy of day-ahead load forecasting and renewable energy power generation forecasting is insufficient, resulting in a significant deviation between the scheduling plan and the actual operation. Specifically, on the one hand, the non-linear characteristics of loads and renewable energy make it difficult for traditional linear forecasting models to capture their complex correlations; on the other hand, the low-precision forecasting makes the capacity configuration and charge-discharge strategies of energy storage systems lack foresight. If the day-ahead plan underestimates the load demand or overestimates the output of renewable energy, the energy storage may prematurely deplete its capacity in the real-time stage and be unable to suppress the power deficit, and vice versa, which will lead to the idleness of the energy storage and reduce its utilization rate. Such deviations not only force the real-time scheduling to frequently activate high-cost backup resources, but also cause battery life attenuation due to the mismatch of energy storage charge-discharge time sequences, ultimately threatening the economic operation and safety stability of the system.

[0004] Therefore, an optimized day-ahead and real-time collaborative scheduling scheme for electric energy substitution loads and energy storage systems is desired. Summary of the Invention

[0005] To solve the above technical problems, this application is proposed. Embodiments of this application provide a day-ahead and real-time collaborative scheduling system for electric energy substitution loads and energy storage systems.

[0006] According to one aspect of this application, a day-ahead and real-time collaborative scheduling system for electric energy substitution loads and energy storage systems is provided, which includes:

[0007] A historical data acquisition module, configured to acquire historical data of electric energy substitution loads and historical data of renewable energy power generation;

[0008] A day-ahead load forecasting module, configured to input historical data of the electric energy substitution load into a load forecasting model to obtain a day-ahead load forecasting result, wherein the day-ahead load forecasting module is configured to: perform spatio-temporal constraint transfer analysis on the historical data of the electric energy substitution load based on a local time-series feature domain to obtain the day-ahead load forecasting result;

[0009] A power generation forecasting module, configured to input historical data of the renewable energy power generation data into a renewable energy power generation forecasting model to obtain a day-ahead renewable energy power generation forecasting result;

[0010] A scheduling plan generation module, configured to generate a day-ahead scheduling plan based on the day-ahead load forecasting result and the day-ahead renewable energy power generation forecasting result and in combination with a preset optimization target;

[0011] An actual data acquisition module, configured to acquire actual electric energy substitution load data and actual renewable energy power generation data;

[0012] A deviation calculation module, configured to calculate deviation information of the actual electric energy substitution load data and the actual renewable energy power generation data based on the day-ahead scheduling plan;

[0013] A power adjustment module, configured to adjust the charge and discharge power of an energy storage system based on the deviation information.

[0014] Compared with the prior art, the day-ahead and real-time coordinated scheduling system for the electric energy substitution load and the energy storage system provided by the present application first acquires historical data of the electric energy substitution load and renewable energy power generation, then inputs them into a load forecasting model and a power generation forecasting model respectively to obtain a day-ahead load forecasting result and a day-ahead power generation forecasting result, and generates a day-ahead scheduling plan based on the obtained forecasting results and a preset optimization target. Subsequently, actual electric energy substitution load and renewable energy power generation data are acquired, and deviation information between the actual data and the forecasting data is calculated according to the day-ahead scheduling plan. Finally, the charge and discharge power of the energy storage system is adjusted based on the deviation information. In this way, the coordinated operation efficiency of the electric energy substitution equipment cluster and renewable energy can be effectively improved, and thus the economic operation and safety stability of the system are effectively ensured. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. In the drawings:

[0016] Figure 1It is a system block diagram of a day-ahead and real-time collaborative scheduling system for an electric energy substitution load and an energy storage system according to an embodiment of the present application.

[0017] Figure 2 It is a block diagram of a day-ahead load forecasting module in a day-ahead and real-time collaborative scheduling system for an electric energy substitution load and an energy storage system according to an embodiment of the present application.

[0018] Figure 3 It is a block diagram of a local time-series encoding unit for an electric energy substitution load in a day-ahead and real-time collaborative scheduling system for an electric energy substitution load and an energy storage system according to an embodiment of the present application.

[0019] Figure 4 It is a block diagram of a local time-series message passing encoding unit for an electric energy substitution load in a day-ahead and real-time collaborative scheduling system for an electric energy substitution load and an energy storage system according to an embodiment of the present application.

[0020] Figure 5 It is a block diagram of a spatio-temporal collaborative constraint factor calculation sub-unit in a day-ahead and real-time collaborative scheduling system for an electric energy substitution load and an energy storage system according to an embodiment of the present application. Detailed implementation manners

[0021] Next, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0022] With the implementation of "dual carbon", the collaborative access of large-scale electric energy substitution loads and high-penetration renewable energy sources poses challenges to the power system in terms of both supply and demand uncertainties. The time-varying characteristics of electric energy substitution loads are affected by multiple factors such as weather, energy consumption behavior, and electricity price policies, while the intermittent output of wind power and photovoltaic power is strongly correlated with meteorological conditions. Both of them jointly exacerbate the difficulty of system dynamic balance. The current dispatching system has twofold constraints: First, traditional linear prediction models are difficult to adapt to the non-linear characteristics of both the power source and load sides, resulting in insufficient accuracy of day-ahead dispatching plans; Second, inaccurate predictions lead to inaccurate energy storage configuration and control strategies, easily triggering polarized risks of "premature depletion of energy storage capacity" or "idle energy storage resources". Such deviations not only increase the real-time dispatching cost of the system, but also accelerate the equipment loss due to the time-series mismatch of energy storage, ultimately endangering the economy and safety margin of the power system.

[0023] Based on this, the present application proposes a day-ahead and real-time collaborative scheduling system for an electric energy substitution load and an energy storage system. Figure 1 It is a system block diagram of a day-ahead and real-time collaborative scheduling system for an electric energy substitution load and an energy storage system according to an embodiment of the present application. As Figure 1As shown in the figure, in the day-ahead and real-time coordinated scheduling system 100 of the electric energy substitution load and the energy storage system, it includes: a historical data acquisition module 110, which is used to acquire the historical data of the electric energy substitution load and the historical data of renewable energy power generation data; a day-ahead load forecasting module 120, which is used to input the historical data of the electric energy substitution load into a load forecasting model to obtain a day-ahead load forecasting result; a power generation forecasting module 130, which is used to input the historical data of the renewable energy power generation data into a renewable energy power generation forecasting model to obtain a day-ahead renewable energy power generation forecasting result; a scheduling plan generation module 140, which is used to generate a day-ahead scheduling plan based on the day-ahead load forecasting result and the day-ahead renewable energy power generation forecasting result and in combination with a preset optimization target; an actual data acquisition module 150, which is used to acquire actual electric energy substitution load data and actual renewable energy power generation data; a deviation calculation module 160, which is used to calculate the deviation information between the actual electric energy substitution load data and the actual renewable energy power generation data based on the day-ahead scheduling plan; and a power adjustment module 170, which is used to adjust the charge and discharge power of the energy storage system based on the deviation information.

[0024] Specifically, in this application, the day-ahead load forecasting module and the power generation forecasting module respectively model the complex characteristics of the electric energy substitution load and renewable energy power generation. This helps to more accurately capture the non-linear relationship between the two, thus alleviating the problem that traditional linear forecasting models are difficult to handle the spatio-temporal coupling effect under extreme weather conditions. In addition, the day-ahead scheduling plan generated by the scheduling plan generation module in combination with the optimization target can, to a certain extent, balance the supply-demand relationship and provide guidance for the capacity configuration and charge-discharge strategy of the energy storage system. Since there are inevitably deviations in actual operation, this application further introduces an actual data acquisition module and a deviation calculation module to monitor in real time the differences between the actual load and power generation conditions and the day-ahead plan. This real-time feedback mechanism enables the power adjustment module to dynamically adjust the charge and discharge power of the energy storage system according to the deviation information, thereby suppressing power deficits or avoiding energy storage idleness in the real-time stage, and reducing the problems of frequent activation of standby resources and energy storage life attenuation caused by prediction errors. This technical solution, through the combination of day-ahead prediction and real-time adjustment, not only improves the accuracy of load and renewable energy power generation forecasting, but also enhances the utilization rate of the energy storage system, thus achieving the precise matching of source-storage-load resources and ultimately ensuring the economic operation and safety stability of the power system.

[0025] In the embodiment of the present application, the historical data acquisition module 110 is configured to acquire the historical data of the electricity substitution load and the historical data of the renewable energy power generation data. It should be understood that the historical data of the electricity substitution load mainly includes the electricity load data of the electricity substitution equipment on the user side at different time points, which can reflect its time-varying characteristics of electricity consumption, such as the electricity load magnitudes in different seasons, different dates (weekdays, holidays, etc.), and different time periods of a day. It covers the information of influencing factors related to electricity consumption behavior, such as the weather conditions at that time (temperature, humidity, wind force, etc., because the weather will affect the power consumption of, for example, electric vehicles for heating or cooling, the operating efficiency of electrolytic hydrogen production equipment, etc.), user habit-related data (such as the charging time period distribution of electric vehicle users, etc.), and the implementation situation of electricity price policies (such as the electricity price levels at different time periods, which will affect users' electricity consumption choices), etc. The historical data of the renewable energy power generation data mainly includes the power generation power information of renewable energy (such as wind power, photovoltaic) at different time points. For wind power, wind speed and wind direction are key influencing factors, while for photovoltaic, meteorological data such as light intensity and sunshine duration are crucial. These meteorological conditions directly determine the power generation capacity of renewable energy, and different meteorological conditions exist at different time points, which will further result in the intermittent and fluctuating characteristics of the power generation power of renewable energy on the time scale. Generally speaking, through the time series analysis of the historical data of the electricity substitution load and the historical data of the renewable energy power generation data, the day-ahead load forecast and the renewable energy power generation forecast can be realized. Based on these two forecast results, a reasonable day-ahead scheduling plan can be generated, so that the power system can reasonably allocate and accurately match resources among the source (renewable energy power generation), storage (energy storage system), and load (electricity substitution load), in order to cope with the challenges of uncertainty on both the supply and demand sides faced by the power system and ensure the stable and economic operation of the system.

[0026] In the embodiment of the present application, the day-ahead load forecasting module 120 is configured to input the historical data of the electricity substitution load into a load forecasting model to obtain a day-ahead load forecasting result. It should be understood that the historical data of the electricity substitution load is a record of the actual load changes under the combined action of various factors over a period of time in the past. The internal laws and patterns of load changes are hidden in these data, such as seasonal change laws, different electricity consumption patterns on weekdays and rest days, the impact of special events or policy adjustments on the load, etc. By analyzing the historical data, the correlation between the load and various influencing factors can be discovered, providing a basis for predicting future loads. However, traditional load forecasting methods are usually based on linear models (such as linear regression, etc.), and these models are difficult to handle the widespread non-linear characteristics and complex interaction patterns in actual load data. For example, under extreme weather conditions, the heating or cooling load of users may increase sharply, while the photovoltaic or wind power generation capacity may drop due to deteriorating meteorological conditions. This spatio-temporal coupling effect is difficult to accurately capture by traditional linear forecasting models.

[0027] Therefore, to address the above technical problems, the technical concept of the present application is to first perform time-series segmentation on the historical load data according to time scales (such as hours, days, weeks) to capture the fluctuation characteristics of local electricity consumption patterns within different periods; then extract the dynamic correlation of load changes in each period through sequence coding; subsequently, adopt message passing coding with spatio-temporal dual constraints to establish the load diffusion correlation between nodes in the spatial dimension and identify the precursor signals of load mutation events in the time dimension; finally, decode the dynamic characteristics across time scales to generate a day-ahead load forecasting result that can reflect complex interactions. This solution significantly improves the load forecasting accuracy in extreme scenarios by explicitly modeling the spatio-temporal non-linear correlation of load changes, making the energy storage charge and discharge strategies in the day-ahead scheduling plan more in line with the actual supply and demand fluctuations, avoiding misconfiguration of energy storage capacity or charge and discharge timing conflicts caused by inaccurate forecasting, thereby reducing the cost of reserve resource invocation in the real-time stage and extending the service life of the energy storage system.

[0028] Specifically, in the embodiment of the present application, the day-ahead load forecasting module is configured to: perform spatio-temporal constraint transfer analysis based on the local time-series feature domain on the historical data of the electricity substitution load to obtain the day-ahead load forecasting result. More specifically, Figure 2 The block diagram of the day-ahead load forecasting module in the day-ahead - real-time coordinated scheduling system of the electricity substitution load and the energy storage system according to the embodiment of the present application. As Figure 2As shown, the day-ahead load forecasting module 120 includes: an electric energy substitution load local time-series encoding unit 121, configured to perform local time-series sequence encoding on the historical data of the electric energy substitution load to obtain a set of electric energy substitution load local time-series correlation feature vectors; an electric energy substitution load local time-series message passing encoding unit 122, configured to perform message passing encoding based on spatio-temporal double constraints on the set of electric energy substitution load local time-series correlation feature vectors to obtain an electric energy substitution load spatio-temporal double-constraint feature vector; and a day-ahead load forecasting result generating unit 123, configured to obtain the day-ahead load forecasting result based on the electric energy substitution load spatio-temporal double-constraint feature vector.

[0029] In the embodiment of the present application, the electric energy substitution load local time-series encoding unit 121 is configured to perform local time-series sequence encoding on the historical data of the electric energy substitution load to obtain a set of electric energy substitution load local time-series correlation feature vectors. Specifically, Figure 3 FIG. is a block diagram of an electric energy substitution load local time-series encoding unit in a day-ahead and real-time collaborative scheduling system for an electric energy substitution load and an energy storage system according to an embodiment of the present application. As Figure 3 shown, the electric energy substitution load local time-series encoding unit 121 includes: an electric energy substitution load data slicing sub-unit 1211, configured to perform data time-series slicing on the historical data of the electric energy substitution load based on a preset time scale to obtain a set of electric energy substitution load local time-series distributions; and an electric energy substitution load local time-series correlation feature extraction sub-unit 1212, configured to perform sequence encoding on each electric energy substitution load local time-series distribution in the set of electric energy substitution load local time-series distributions to obtain the set of electric energy substitution load local time-series correlation feature vectors.

[0030] In the embodiment of the present application, the power substitution load data sub - division unit 1211 is configured to perform data time - series segmentation on the historical data of the power substitution load based on a preset time scale to obtain a set of local time - series distributions of the power substitution load. Correspondingly, considering that the power substitution load data has complex time - series characteristics, including variation rules at multiple time scales. Different time scales (such as hours, days, weeks, etc.) correspond to different electricity consumption behavior patterns and influencing factors. For example, taking the hour as the scale can capture the electricity consumption peaks and valleys at different times of a day; taking the day as the scale can reflect the electricity consumption differences between weekdays and weekends; taking the week as the scale helps to discover the weekly electricity consumption cycle rules. Based on this, in the present application, data time - series segmentation is performed on the historical data of the power substitution load based on a preset time scale to obtain a set of local time - series distributions of the power substitution load. In this way, the original historical data can be sorted according to different time granularities, which is convenient for more in - depth exploration of various rules and characteristics hidden in the data. That is, after the historical data of the power substitution load is segmented into a set of local time - series distributions, refined analysis can be performed on the load data within each local time period, and it is possible to more clearly observe the fluctuation characteristics, change trends, and outliers of the load at different time scales. For example, it can be analyzed that the rules for the occurrence of load peaks and valleys at different times of the day (morning, noon, and evening) or the load change patterns on specific days of the week, so as to better understand the internal mechanism of load changes and provide more accurate feature information for subsequent prediction models.

[0031] In the embodiment of the present application, the local time - series correlation feature extraction sub - unit 1212 of the electricity - substitution load is used to perform sequence coding on each local time - series distribution of the electricity - substitution load in the set of local time - series distributions of the electricity - substitution load to obtain a set of local time - series correlation feature vectors of the electricity - substitution load. Specifically, in the embodiment of the present application, the local time - series correlation feature extraction sub - unit of the electricity - substitution load is used to: perform sequence coding based on one - dimensional convolution on each local time - series distribution of the electricity - substitution load in the set of local time - series distributions of the electricity - substitution load to obtain a set of local time - series correlation feature vectors of the electricity - substitution load. Correspondingly, considering that the load change not only reflects the numerical fluctuation in the time series, but also contains a non - linear correlation pattern of multi - factor dynamic interaction. For example, the load at the current moment may have a certain correlation with the loads at multiple past moments and the future load trend. Traditional methods directly input the original load time series into the prediction model, making it difficult to effectively distinguish the dominant factors and secondary factors in such dynamic correlations. Therefore, in the technical solution of the present application, sequence coding is performed on each local time - series distribution of the electricity - substitution load in the set of local time - series distributions of the electricity - substitution load to effectively capture the dynamic changes and internal correlations of the load data in the time dimension, obtaining a set of local time - series correlation feature vectors of the electricity - substitution load. In particular, in a specific example of the present application, sequence coding based on one - dimensional convolution is performed on each local time - series distribution of the electricity - substitution load in the set of local time - series distributions of the electricity - substitution load to obtain a set of local time - series correlation feature vectors of the electricity - substitution load. It can be understood that the one - dimensional convolution operation can automatically extract local features in the time series. For electricity - substitution load data, there are local correlations and patterns between the load values at different time points. For example, during certain periods of a day, the load may show specific change trends, such as the electricity peak after getting up in the morning and the electricity trough before going to bed at night. One - dimensional convolution can effectively capture these local features through convolution operations within a local range by sliding a window, and transform them into more representative feature representations.

[0032] In the embodiment of the present application, the local time - series message - passing coding unit 122 of the electricity - substitution load is used to perform message - passing coding based on spatio - temporal dual constraints on the set of local time - series correlation feature vectors of the electricity - substitution load to obtain spatio - temporal dual - constraint feature vectors of the electricity - substitution load. Specifically, Figure 4 The block diagram of the local time - series message - passing coding unit of the electricity - substitution load in the day - ahead - real - time coordinated scheduling system of the electricity - substitution load and the energy storage system according to the embodiment of the present application. As Figure 4As shown, the local time-series message passing encoding unit 122 of the electricity substitution load includes: a time-series feature sequence encoding subunit 122-1, configured to input the set of local time-series correlation feature vectors of the electricity substitution load into a sequence encoder based on a recurrent neural network to obtain a set of initial encoding vectors of the local time-series correlation features of the sequence-transmitted electricity substitution load; a spatio-temporal collaborative constraint factor calculation subunit 122-2, configured to calculate the electricity substitution load local time-series correlation feature message passing spatio-temporal collaborative constraint factors of each initial encoding vector of the local time-series correlation features of the sequence-transmitted electricity substitution load in the set; and an electricity substitution load spatio-temporal aggregation subunit 122-3, configured to perform message passing structure modulation aggregation on each initial encoding vector of the local time-series correlation features of the sequence-transmitted electricity substitution load based on the electricity substitution load local time-series correlation feature message passing spatio-temporal collaborative constraint factors to obtain the electricity substitution load spatio-temporal double-constraint feature vector.

[0033] It should be understood that there are multi-dimensional dynamic interaction features hidden in the local time-series fluctuations of the electricity substitution load, which not only show periodic or mutant patterns in the time dimension, but also form a complex coupling relationship with the spatial distribution. For example, the start-stop operation of the electrolytic hydrogen equipment in the industrial park will generate a steep load step in the hourly time series, but the amplitude and duration of this step are subject to the real-time constraints of production plan adjustment and grid node voltage, forming non-linear features with spatio-temporal dependence. Traditional sequence encoding methods (such as simple moving average or Fourier transform) can only extract the macroscopic statistical characteristics of the load curve and cannot capture the microscopic dynamics of load mutation events in local time periods (such as details like the rising edge slope and peak maintenance duration of load pulses). This shallowness of feature expression leads to a lack of in-depth analysis of the load evolution mechanism in subsequent spatio-temporal modeling. Especially when new electricity substitution equipment is connected in load-intensive areas, the prediction model is prone to deviation due to feature loss. Based on this, the present application performs message passing encoding with dual spatio-temporal constraints on the set of local time-series correlation feature vectors of the electricity substitution load to obtain the electricity substitution load spatio-temporal double-constraint feature vector.

[0034] Specifically, in the embodiment of the present application, the time-series feature sequence encoding subunit 122-1 is configured to: input the set of local time-series correlation feature vectors of the electricity substitution load into a sequence encoder based on a recurrent neural network to obtain a set of initial encoding vectors of the local time-series correlation features of the sequence-transmitted electricity substitution load, which can be represented by the following formula:

[0035] I = {v1, v2,..., v i ,..., v t}

[0036] RNN(I) = {h1, h2,..., h i ,..., h t}

[0037] Wherein, I is a set of local time series correlation feature vectors of the electric energy substitution load, v1, v2, v i and v t are respectively the 1st, 2nd, i-th, and t-th flow local time series hidden mode feature coding vectors in the set of local time series correlation feature vectors of the electric energy substitution load. RNN(I) is the sequence coding of I based on the RNN structure, and h1, h2, h i and h t are respectively the 1st, 2nd, i-th, and t-th sequence transfer initial coding vectors of the local time series correlation features of the electric energy substitution load in the set of initial coding vectors of the local time series correlation features of the electric energy substitution load passed by the sequence.

[0038] It should be understood that traditional moving average or Fourier transform can only extract the macroscopic statistical characteristics of the load curve and it is difficult to capture microscopic dynamic features such as load steps caused by the start and stop of electrolytic hydrogen production equipment. The Recurrent Neural Network (RNN) can model the dynamic dependencies in the time dimension through its inherent time series memory ability, especially for capturing the time series patterns of mutation events (such as steep rising edges, peak maintenance duration) within a local time period. Its recurrent connection structure enables historical information to be transmitted and accumulated between time steps, thereby initially encoding the periodicity, mutability, and complex coupling relationships in the time dimension of the load. That is, the generated set of initial coding vectors of the local time series correlation features of the electric energy substitution load passed by the sequence not only retains the local time series fluctuation details of the original load data (such as hourly load steps), but also provides a preliminary time series feature basis for the introduction of subsequent spatio-temporal constraints, solving the problem of shallow feature expression in traditional methods.

[0039] Specifically, Figure 5 is a block diagram of the spatio-temporal coordination constraint factor calculation sub-unit in the day-ahead and real-time coordinated scheduling system of the electric energy substitution load and energy storage system according to the embodiment of the present application. As Figure 5As shown, the spatio-temporal collaborative constraint factor calculation subunit 122-2 includes: a temporal confidence constraint factor calculation secondary subunit 122-21, configured to calculate the temporal confidence constraint factor of the local temporal correlation feature of the electricity substitution load for each initial coding vector of the local temporal correlation feature of the sequence transmission electricity substitution load in the set of initial coding vectors of the local temporal correlation feature of the sequence transmission electricity substitution load; a spatial confidence constraint factor calculation secondary subunit 122-22, configured to calculate the spatial confidence constraint factor of the local temporal correlation feature of the electricity substitution load for each initial coding vector of the local temporal correlation feature of the sequence transmission electricity substitution load in the set of initial coding vectors of the local temporal correlation feature of the sequence transmission electricity substitution load; and a message passing spatio-temporal collaborative constraint factor construction secondary subunit 122-23, configured to construct the message passing spatio-temporal collaborative constraint factor of the local temporal correlation feature of the electricity substitution load for each initial coding vector of the local temporal correlation feature of the sequence transmission electricity substitution load based on the spatial confidence constraint factor and the temporal confidence constraint factor of the local temporal correlation feature of the electricity substitution load for each initial coding vector of the local temporal correlation feature of the sequence transmission electricity substitution load.

[0040] More specifically, in the embodiment of the present application, the temporal confidence constraint factor calculation secondary subunit 122-21 is configured to: calculate the temporal confidence constraint factor of the local temporal correlation feature of the electricity substitution load for each initial coding vector of the local temporal correlation feature of the sequence transmission electricity substitution load in the set of initial coding vectors of the local temporal correlation feature of the sequence transmission electricity substitution load, which can be expressed by the following formula:

[0041]

[0042] Where, W 1i is the learnable weight matrix corresponding to h i W 2i is the learnable weight matrix corresponding to v i α and β are respectively trainable weighted hyperparameters, tanh is the hyperbolic tangent activation function, is matrix multiplication, v t is the temporal scoring weight vector, is the local temporal energy score of the electricity substitution load corresponding to h i softmax is the normalization function, is the temporal confidence constraint factor of the local temporal correlation feature of the electricity substitution load corresponding to h i

[0043] It should be understood that the contribution of load mutation events to the overall prediction varies significantly at different time steps. For example, the sudden increase in the load of power-to-hydrogen during the peak photovoltaic generation period at noon (triggering centralized production due to the low electricity price valley) has strong indicative significance for the prediction of the evening peak load, while the random fluctuations during the late-night low-load period may belong to noise. By calculating the temporal confidence constraint factor of the local temporal correlation characteristics of the power-to-energy substitution load through the attention mechanism, the model can dynamically identify the key time steps. Specifically, when the initial encoding vector at a certain moment shows a sudden increase in load accompanied by a jump in the electricity price signal, the attention weight will significantly increase the confidence of this vector. Conversely, if the load fluctuation has no external factor correlation, the weight will be reduced. This mechanism simulates the selective attention of humans to temporal information. For example, it preferentially processes load mutation events synchronized with electricity price policies while suppressing the influence of isolated noise.

[0044] More specifically, in the embodiment of the present application, the spatial confidence constraint factor calculation secondary subunit 122-22 is used to: calculate the power-to-energy substitution load local temporal correlation feature spatial confidence constraint factor of each sequence transfer power-to-energy substitution load local temporal correlation feature initial encoding vector in the set of sequence transfer power-to-energy substitution load local temporal correlation feature initial encoding vectors, which can be expressed by the following formula:

[0045]

[0046] where h i T is the transposed vector of h i , is the square of the calculation of the F norm, s(h i ) is the power-to-energy substitution load local temporal correlation feature spatial similarity score value corresponding to h i , exp is the exponential function value with the natural constant e as the base, is the power-to-energy substitution load local temporal correlation feature spatial confidence constraint factor corresponding to h i .

[0047] It should be understood that load fluctuations not only have temporal locality but are also strongly correlated with the spatial location in the power grid topology. For example, a load mutation at the core distribution node of an industrial park may trigger a cascading effect, which requires a higher structural importance to be assigned. By explicitly modeling the spatial structure of the node feature distribution, its topological role can be quantified. That is, the generated power-to-energy substitution load local temporal correlation feature spatial confidence constraint factor transforms the power grid topology constraint into a spatial weight, enabling the encoding vectors in the load-intensive area or key equipment nodes to obtain a higher propagation priority. For example, if the power-to-hydrogen equipment is located at a multi-equipment parallel node, its spatial confidence is increased to ensure that its start-stop event is not submerged by the noise of neighboring nodes during message passing.

[0048] More specifically, in the embodiments of the present application, the message-passing spatio-temporal collaborative constraint factor constructs a secondary subunit 122-23 for: constructing the message-passing spatio-temporal collaborative constraint factor of the local temporal correlation feature of the electricity substitution load for each sequence to transmit the initial coding vector of the local temporal correlation feature of the electricity substitution load, based on the spatio-temporal confidence constraint factor of the local temporal correlation feature of the electricity substitution load and the temporal confidence constraint factor of the local temporal correlation feature of the electricity substitution load for each sequence to transmit the initial coding vector of the local temporal correlation feature of the electricity substitution load. It can be expressed by the following formula:

[0049]

[0050] where ω1 and ω2 are respectively and contribution adjustment parameters of, sigmoid is the sigmoid function, is the message-passing spatio-temporal collaborative constraint factor of the local temporal correlation feature of the electricity substitution load corresponding to h i

[0051] It should be understood that the spatio-temporal coupling of the load is reflected in that: the mutation in the time dimension needs to satisfy the grid constraints in the space dimension (for example, the step amplitude of the load at a certain node is limited by the voltage stability of adjacent nodes). By adopting a fusion strategy (spatio-temporal constraint "logical AND"), it can be ensured that information is fully transmitted only when the time importance and the space importance are both satisfied. For example, if the load step at a certain moment exceeds the voltage tolerance of the associated node, even if the time confidence is high, its comprehensive weight will still be suppressed by the space constraint. That is, the message-passing spatio-temporal collaborative constraint factor of the local temporal correlation feature of the electricity substitution load realizes the "spatio-temporal gating" function, solving the overfitting problem caused by traditional single-dimensional weighting (such as only focusing on time mutation and ignoring the grid safety boundary). Through this non-linear collaborative effect, it is possible to accurately screen out the feature patterns that are both temporally significant and conform to the space structure constraints.

[0052] Preferably, in another example of the present application, the message-passing spatio-temporal collaborative constraint factor constructs a secondary subunit for: performing spatio-temporal confidence constraint collaborative optimization based on curvature compensation on the spatio-temporal confidence constraint factor of the local temporal correlation feature of the electricity substitution load and the temporal confidence constraint factor of the local temporal correlation feature of the electricity substitution load for each sequence to transmit the initial coding vector of the local temporal correlation feature of the electricity substitution load, so as to obtain the message-passing spatio-temporal collaborative constraint factor of the local temporal correlation feature of the electricity substitution load for each sequence to transmit the initial coding vector of the local temporal correlation feature of the electricity substitution load.

[0053] That is, in the temporal confidence constraint factor of the local temporal correlation feature of the electricity substitution load and the spatio-temporal confidence constraint factor of the local temporal correlation feature of the electricity substitution load​ respectively used to represent the initial encoding vectors h of the local temporal correlation features of the power substitution load for each sequence to transmit electric energy i In the case of the information weight distribution in the time dimension and the space dimension, if the time dimension and the space dimension are separately regarded as a single-degree-of-freedom mode under the spatio-temporal global, it is expected to avoid the negative curvature representation of the spatio-temporal joint space caused by the negative attraction related to the single-mode attention, resulting in a reduction in the fusion space expression effect under non-planar conditions.

[0054] Therefore, first, based on the initial encoding vectors h of the local temporal correlation features of the power substitution load for each sequence to transmit electric energy i the corresponding temporal confidence constraint factor of the local temporal correlation features of the power substitution load and the spatial confidence constraint factor of the local temporal correlation features of the power substitution load to construct the constant curvature space representation G i and the spherical coordinate approximation representation R i :

[0055]

[0056] That is, when it can make the single-mode correlation representations and in the time and space dimensions approach the flat spatio-temporal coupling attraction, thereby reflecting the approximate Euclidean property within the fusion space. In this way, the spatio-temporal collaborative constraint factor for message passing of the local temporal correlation features of the power substitution load can fuse the space metric reference representation, that is, by compensating for the generation of single-mode negative curvature to achieve the plane preservation within the fusion space, thereby enhancing the expression effect of the spatio-temporal collaborative constraint factor for message passing of the local temporal correlation features of the power substitution load

[0057] Specifically, in the embodiment of the present application, the spatio-temporal aggregation sub-unit 122-3 of the power substitution load is used to: based on the spatio-temporal collaborative constraint factor for message passing of the local temporal correlation features of the power substitution load, perform message passing structure modulation on the initial encoding vectors of the local temporal correlation features of the power substitution load for each sequence to obtain a set of structurally modulated encoding vectors of the local temporal correlation features of the power substitution load for sequence transmission, and this process can be represented by the formula:

[0058]

[0059] where wh i is the i-th structurally modulated encoding vector of the local temporal correlation features of the power substitution load for sequence transmission in the set of structurally modulated encoding vectors of the local temporal correlation features of the power substitution load for sequence transmission;

[0060] ​Calculate the position-wise sum of the set of structural modulation coding vectors of the local temporal correlation features of the sequence of transferred electricity for the electricity substitution load to obtain the spatio-temporal double-constraint feature vector of the electricity substitution load. This process can be expressed by the formula:

[0061]

[0062] where t is the number of vectors in the set of structural modulation coding vectors of the local temporal correlation features of the sequence of transferred electricity for the electricity substitution load, and z is the spatio-temporal double-constraint feature vector of the electricity substitution load.

[0063] Correspondingly, the access of electricity substitution equipment may trigger the spatial propagation of local load pulses (such as the power oscillation of the feeder caused by the start-stop of the electric boiler group in a certain park). By dynamically adjusting the message passing path and intensity through the spatio-temporal collaborative constraint factor of the local temporal correlation feature message of the electricity substitution load, the organic propagation process of load mutations in the power grid can be simulated. For example, for coding vectors with high spatio-temporal confidence, enhance their propagation intensity between adjacent nodes; for vectors with low confidence, limit their diffusion range. That is, the generated set of structural modulation coding vectors of the local temporal correlation features of the sequence of transferred electricity for the electricity substitution load can characterize the propagation path of load pulses (such as the attenuation process from the hydrogen production equipment to the distribution main line), solving the problem of insufficient modeling of spatio-temporal coupling relationships in traditional models. Especially in the scenario of new equipment access, it can avoid prediction deviations caused by the lack of consideration of spatial constraints for local mutations.

[0064] It should be understood that load forecasting needs to transform dynamic spatio-temporal features into decision inputs of a fixed dimension. The position-wise sum operation compresses the time dimension of the set of structural modulation coding vectors of the local temporal correlation features of the sequence of transferred electricity for the electricity substitution load, retaining the statistical significance of key spatio-temporal patterns (such as the cumulative energy of step events and the spatial propagation range). Compared with simple pooling, its advantage lies in retaining the modulation weight information of each time step. That is, the finally obtained spatio-temporal double-constraint feature vector of the electricity substitution load contains both the microscopic dynamics of local mutations (such as the time evolution of the rising edge slope) and the propagation constraints in the power grid topology (such as the amplitude attenuation under node voltage limitations), and can provide a feature expression with both fine-grainedness and structural security for subsequent load forecasting.

[0065] In the embodiment of the present application, the day-ahead load prediction result generation unit 123 is configured to obtain the day-ahead load prediction result based on the spatio-temporal dual-constraint feature vector of the electricity substitution load. Specifically, in the embodiment of the present application, the day-ahead load prediction result generation unit is configured to: perform feature decoding on the spatio-temporal dual-constraint feature vector of the electricity substitution load based on the RNN model to obtain the day-ahead load prediction result. It should be understood that the electricity substitution load data is essentially sequential data with a chronological order, and there is a dependency relationship between its values at different time points. For example, the load at the current moment is often affected by the load change trend in the past period. The RNN model is specifically designed to process sequential data. It has a cyclic structure and can retain the previous information when processing the current input, that is, it has a memory function. This enables the RNN to well capture the time series features and long-term dependency relationships in the electricity substitution load data, thereby making a more accurate prediction of the future value of the load. Therefore, through the processing and learning of the spatio-temporal features by the RNN model, the time and space information in the historical load data can be fully utilized, considering the long-term dependency relationships and dynamic characteristics of the load change, so as to provide reliable load prediction data for the day-ahead scheduling of the power system. Accurate prediction results help to reasonably arrange the power generation plan, the charge and discharge strategies of the energy storage system, and other relevant resources, improving the operation efficiency and economy of the power system. In detail, the day-ahead load prediction result is essentially a refined spatio-temporal coupling prediction of the electricity substitution load demand in the next 24 hours. The prediction result is specifically manifested as the predicted load power values for each future time period (usually with a time granularity of 15 minutes or 30 minutes), such as the concentrated operation power peak of the electrolytic hydrogen production equipment during the peak photovoltaic power generation period at noon. In particular, the day-ahead load prediction result also outputs the classified load prediction results for different electricity substitution scenarios such as electric vehicles, electrolytic hydrogen production, and electric heating. For example, the prediction result can distinguish the proportion of the charging power of electric vehicles during the evening peak period and the load demand of industrial electric heating under a specific production plan, which can provide a basis for subsequent refined scheduling.

[0066] In summary, the day-ahead load forecasting module 120 has been clearly described. First, it temporally segments the historical load data according to time scales (such as hours, days, weeks) to capture the fluctuation characteristics of local electricity consumption patterns in different cycles. Then, it extracts the dynamic correlation of load changes in each time period through sequence encoding. Subsequently, it uses message passing encoding with spatio-temporal dual constraints to establish the load diffusion correlation between nodes in the spatial dimension and identify the precursor signals of load mutation events in the time dimension. Finally, it decodes the dynamic characteristics across time scales to generate a day-ahead load forecasting result that can reflect complex interactions. In this way, by explicitly modeling the spatio-temporal non-linear correlation of load changes, the load forecasting accuracy in extreme scenarios can be significantly improved, thereby making the energy storage charge and discharge strategies in the day-ahead scheduling plan more in line with the actual supply and demand fluctuations, avoiding misconfiguration of energy storage capacity or charge and discharge timing conflicts caused by inaccurate forecasting, and further reducing the cost of reserve resource utilization in the real-time stage and extending the service life of the energy storage system.

[0067] In the embodiment of the present application, the power generation prediction module 130 is configured to input historical data of the renewable energy power generation data into a renewable energy power generation prediction model to obtain a day-ahead renewable energy power generation prediction result. It should be understood that the power generation of renewable energy (such as wind power, photovoltaic power, etc.) has significant intermittency, volatility, and uncertainty. Its power generation power is closely related to meteorological conditions (such as wind speed, light intensity, temperature, etc.). For example, wind power generation depends on the magnitude of the wind speed, and photovoltaic power generation depends on the light intensity and sunshine duration. These meteorological factors vary erratically in different times and spaces, making it difficult to accurately estimate the renewable energy power generation power. By analyzing the historical data of renewable energy power generation data, the internal relationship and variation law between the power generation power and meteorological conditions and other factors can be mined, providing a data basis for prediction. In particular, in a specific example of the present application, first, the historical data of renewable energy power generation data is collected and sorted out. These data cover the actual power generation records of renewable energy power generation equipment (such as wind turbines, photovoltaic panels, etc.) over a long period of time in the past. At the same time, comprehensive multi-dimensional meteorological data related to it is collected. For wind power generation, data such as wind speed, wind direction, and air density need to be focused on, and for photovoltaic power generation, meteorological information such as irradiance, temperature, and cloud thickness needs to be mainly collected. After the collection is completed, these data are subjected to spatio-temporal alignment processing. This step is extremely crucial and it is necessary to ensure that the power generation data and meteorological data are strictly corresponding in terms of time stamps and spatial positions to ensure data consistency. For example, for multiple wind farms distributed in different geographical locations, according to their specific longitude and latitude information, the power generation data of each wind farm should be accurately matched with the meteorological data of the corresponding region and corresponding moment to extract the accurate meteorological time series characteristics of the regions where each wind farm is located. Then, a model suitable for renewable energy power generation prediction is constructed. Here, a model combining a graph neural network and an attention mechanism is taken as an example. Different power generation equipment (such as each wind turbine or photovoltaic power station) is regarded as a node in the model, and then these nodes are embedded into the power grid topology structure. By simulating the actual situation of power transmission in the power grid and using the message passing mechanism inside the model, the effects generated by the propagation of meteorological factors between different regions can be effectively captured. For example, when a certain region is affected by strong winds and the power generation power of the wind farm increases, this meteorological change may affect adjacent regions along with the atmospheric circulation, causing the power generation power of the wind farms in adjacent regions to also fluctuate accordingly. The model can simulate and predict this chain reaction through this mechanism. During the operation of the model, with the help of the spatio-temporal attention mechanism, weights are dynamically assigned to the impacts of different meteorological factors on power generation output. For example, in sunny weather, the influence weight of irradiance on photovoltaic power generation power is relatively large, and the model will automatically enhance the role of the meteorological factor of irradiance in the prediction process; while in rainy weather, meteorological factors such as cloud thickness and precipitation intensity have a significant impact on photovoltaic power generation power, and the model will correspondingly increase the weights of these factors and give priority to considering their impacts on power generation output.After a series of complex operations and processes of the model, the finally output is the current renewable energy power generation prediction result. This result presents the predicted values of the renewable energy power generation at different times within the next day, fully reflecting the changing trend of renewable energy power generation under the interaction of meteorological evolution and power grid structure, and can provide a very accurate prediction of the renewable energy output trend for subsequent energy storage scheduling and overall energy distribution in the power system, helping the power system to operate more efficiently and stably.

[0068] In the embodiment of the present application, the scheduling plan generation module 140 is configured to generate a day-ahead scheduling plan based on the day-ahead load prediction result and the day-ahead renewable energy power generation prediction result and in combination with a preset optimization goal. Correspondingly, considering that the day-ahead load prediction result reflects the change situation of the power load within the next day, including information such as the magnitude of the load, the change trend, and the possible peak and trough periods. The day-ahead renewable energy power generation prediction result provides the predicted situation of renewable energy power generation, such as the magnitude of the power generation and the fluctuation range. Combining these two prediction results can comprehensively understand the supply and demand situation of the power system. In addition, the preset optimization goal (such as cost minimization, carbon emission reduction, system stability improvement, etc.) represents the desired direction of the power system operation. Considering these information comprehensively can make the scheduling plan more scientifically, make full use of various resources, and realize the optimized operation of the power system.

[0069] Specifically, when generating the day-ahead scheduling plan, comprehensively and deeply considering the day-ahead load forecast results and the day-ahead renewable energy generation forecast results is the top priority. For the day-ahead load forecast results, they are the hourly load demand values for the next 24 hours obtained by using advanced forecasting models based on historical load data, combined with multiple factors such as weather conditions, user behavior habits, and special events. These load demands cover different types such as residential electricity consumption, commercial electricity consumption, and industrial electricity consumption, and each type of electricity consumption has its unique variation pattern at different time periods. For example, residential electricity consumption usually peaks in the morning and evening because these are the concentrated periods of residents' daily electricity consumption; commercial electricity consumption is related to the business hours of shopping malls and office buildings, and the daytime on weekdays is the peak electricity consumption period; industrial electricity consumption is relatively stable, but it is also affected by production plans and equipment operation cycles. The day-ahead renewable energy generation forecast results are the power generation powers of different renewable energies (such as wind power and photovoltaic) within the next 24 hours obtained based on the historical power generation data of renewable energy generation equipment, meteorological condition forecasts, etc. For wind power, its power generation power mainly depends on meteorological factors such as wind speed, wind direction, and air density, and due to differences in terrain and climate conditions, the power generation characteristics of wind farms in different geographical locations are also different. For example, wind farms in coastal areas may be affected by sea breezes in summer, with relatively stable and high power generation; while wind farms in inland mountainous areas may be affected by valley winds, with large fluctuations in power generation. For photovoltaic, its power generation power is mainly affected by factors such as light intensity, sunshine duration, and temperature, and the power generation amount of photovoltaic power generation varies significantly under different seasons and weather conditions. For example, the photovoltaic power generation power is high on sunny days, while it will decrease significantly on cloudy or rainy days.

[0070] Next, the resource allocation work is carried out according to the preset optimization goals. The preset optimization goals are determined according to the actual situation and development needs of the power system. Common optimization goals include minimizing operating costs, maximizing the utilization rate of renewable energy, minimizing carbon emissions, etc. If the optimization goal is to minimize operating costs, then when allocating resources, the system will give priority to using energy with lower costs. During the low electricity price period, the system will increase the proportion of traditional energy generation, and arrange for the energy storage system to charge to store low-priced electricity; during the peak electricity price period, reduce traditional energy generation, increase the discharge of the energy storage system, and make full use of renewable energy generation to reduce the cost of purchasing electricity from the power grid. If the goal is to maximize the utilization rate of renewable energy, the system will try to give priority to renewable energy generation. When renewable energy generation is sufficient, reduce the investment in traditional energy generation, and even store excess electricity in the energy storage system. When renewable energy generation is insufficient, traditional energy generation and energy storage system discharge are reasonably allocated to meet load demand. If the goal is to minimize carbon emissions, the system will significantly increase the proportion of renewable energy use, reduce dependence on traditional energy with high carbon emissions (such as coal-fired power generation), and ensure that carbon emissions can be reduced even when renewable energy generation fluctuates through the regulation of the energy storage system.

[0071] In order to achieve these optimization goals, advanced mathematical models and algorithms need to be used. Linear programming is a commonly used method that can find the optimal solution of the objective function under a series of linear constraints. In power dispatching, linear programming can be used to determine the power distribution of different energy generation equipment to achieve goals such as minimizing costs or maximizing the utilization of renewable energy. For example, by establishing a linear programming model, the power generation cost, upper and lower limits of power generation, load demand, etc. are used as constraints to solve the optimal power generation of each power generation equipment in each period. Mixed integer programming considers both continuous variables (such as power generation) and discrete variables (such as the start and stop status of power generation equipment), which is suitable for more complex dispatching problems. For some large power systems, which contain multiple types of power generation equipment, the start and stop of some equipment needs to consider factors such as fixed costs and start-up time. Mixed integer programming can simulate this situation more accurately and obtain a better dispatching solution. Machine learning algorithms, such as neural networks and genetic algorithms, can also handle highly nonlinear problems. By learning a large amount of historical data and mining the potential laws in the data, they can more accurately predict load and renewable energy generation and optimize dispatching strategies.

[0072] During the resource allocation process, the system needs to comprehensively consider the balance between power supply and demand, and at the same time deeply analyze various factors affecting the system operation. Weather forecast is one of the key factors. It not only affects the power generation capacity of renewable energy, but also the load demand. For example, in a hot summer, if the weather forecast shows high temperature for the next day, the air-conditioning electricity demand of residents and commercial premises will increase significantly. At the same time, high temperature may reduce the efficiency of photovoltaic power generation equipment. Therefore, the system needs to adjust the dispatching plan in advance according to the weather forecast, increase the standby capacity of traditional energy generation or energy storage systems to cope with possible power shortages. Market electricity price fluctuations are also an important factor. The price differences in different time periods will affect the operation cost of the power system. The system needs to monitor the changes in market electricity prices in real time, increase the charging volume of the energy storage system during the low electricity price period, and reduce the electricity purchase from the grid during the high electricity price period, and reduce costs through reasonable charging and discharging strategies. Grid transmission limitations cannot be ignored either. The transmission lines in the power system have their maximum transmission capacity limitations. If the load demand in a certain area is too large and the transmission capacity of the transmission line is limited, it may lead to power supply shortages in that area. Therefore, the system needs to reasonably allocate generation resources according to the grid transmission limitations to avoid overload phenomena.

[0073] To improve the stability and reliability of the power system, a certain amount of standby capacity is also set in the dispatching plan. Standby capacity refers to the generation capacity reserved outside the normal power generation plan to cope with sudden load increases or power generation equipment failures. The size of the standby capacity is usually determined according to factors such as the load characteristics of the system, the reliability of the power generation equipment, and historical failure data. Generally speaking, for a system with large load fluctuations and low reliability of power generation equipment, a larger standby capacity needs to be set. At the same time, an emergency response plan for emergencies is also formulated. Emergencies may include natural disasters (such as earthquakes, typhoons, heavy rains, etc.), equipment failures (such as generator failures, transmission line failures, etc.) and human sabotage. For different types of emergencies, the emergency response plan will have different countermeasures. For example, when some power generation equipment is damaged due to natural disasters, the system will quickly start the standby power generation equipment and at the same time adjust the operation mode of the transmission line to ensure the power supply to important users; when an equipment failure occurs, the system will promptly isolate the faulty equipment and switch to the standby equipment to ensure the normal operation of the power system. By setting the standby capacity and formulating the emergency response plan, the power system can still operate stably and reliably in the face of various uncertainties, meet the electricity demand of users, and at the same time maximize the dual benefits of economy and environmental protection.

[0074] In the embodiment of the present application, the actual data acquisition module 150 is configured to acquire actual electric energy substitution load data and actual renewable energy power generation data. It should be understood that by acquiring the actual electric energy substitution load data and the actual renewable energy power generation data, the system can accurately grasp the actual level of the current electric energy substitution load and the true output of the renewable energy power generation. Furthermore, the deviation between the prediction and the actual situation can be timely discovered, so as to provide a basis for adjusting the charge and discharge power of the energy storage system based on the deviation information in the subsequent stage, enabling the energy storage system to better play a regulatory role, balance the power supply and demand, and ensure the stable operation of the power system.

[0075] In the embodiment of the present application, the deviation calculation module 160 is configured to calculate deviation information of the actual electric energy substitution load data and the actual renewable energy power generation data based on the day-ahead scheduling plan. It should be understood that the day-ahead scheduling plan is formulated based on prediction data and preset goals. In actual operation, affected by various factors, such as the impact of weather changes on renewable energy power generation and the uncertainty of user electricity consumption behavior, the actual electric energy substitution load data and the actual renewable energy power generation data often differ from the plan. Calculating the deviation information can clarify the degree of this difference. In particular, in a specific example of the present application, when calculating the deviation information of the actual electric energy substitution load data and the actual renewable energy power generation data based on the day-ahead scheduling plan, the system will accurately collect the actual electric energy substitution load data and the actual renewable energy power generation data at precise time scales, such as every 15 minutes or every 30 minutes. The collected actual electric energy substitution load data is carefully compared with the predicted electric energy substitution load value at the same time interval in the day-ahead scheduling plan according to the corresponding time point. Calculate the difference between the two. If the actual load is P1 and the predicted load is P2, the difference ΔP = P1 - P2, and this difference is the deviation basic data of the electric energy substitution load in this time period. For the renewable energy power generation data, similarly according to the corresponding time point, the actual power generation is compared with the predicted power generation in the day-ahead scheduling plan. Assuming the actual power generation power is E1 and the predicted power generation power is E2, the difference ΔE = E1 - E2, and this difference is the deviation basic data of the renewable energy power generation in this time period.

[0076] Considering the complexity of power system operation, different types of deviations have different degrees of impact on the system. The system sets corresponding weight coefficients for different types of deviations according to the impact degree of the deviations on aspects such as system stability and economy. For example, during the peak electricity consumption period, the positive deviation of the electricity substitution load has a greater impact on the system power supply pressure. At this time, a higher weight coefficient w1 is assigned to the positive deviation of the electricity substitution load during this period; while during the period when renewable energy generation is abundant, if the negative deviation of renewable energy generation has a smaller impact on the system's accommodation capacity, a lower weight coefficient w2 can be assigned to it. By multiplying the basic deviation data by the corresponding weight coefficient, the weighted deviation data is obtained. For example, the weighted deviation of the electricity substitution load is ΔP×w1, and the weighted deviation of renewable energy generation is ΔE×w2.

[0077] Over time, the system continuously collects and calculates the deviation data for each period and records these data in sequence according to the time order. After a certain time period, such as one day or one week, the deviation data for all recorded periods is sorted and analyzed. By plotting the curve of the deviation changing with time, the deviation between the actual operation and the day-ahead scheduling plan in different periods is visually displayed. At the same time, data characteristics such as the maximum value, minimum value, and average value of the deviation are statistically calculated to further quantify the overall deviation degree. For example, the average value of the deviation of the electricity substitution load within one day is calculated as Σ(ΔP×w1) / n (n is the number of statistical periods), and the maximum value of the deviation of renewable energy generation within one week is Max(ΔE×w2).

[0078] Finally, the deviation information after sorting and analysis, including the detailed deviation data for each period, deviation statistical characteristics, and deviation change curves, etc., is provided to the subsequent power adjustment module comprehensively and accurately. These deviation information become the key basis for optimizing the charge and discharge strategy of the energy storage system and dynamically adjusting the real-time scheduling decision, so as to ensure that the power system can still maintain good economy and stability in the face of complex and changing actual operation conditions.

[0079] In the embodiment of the present application, the power adjustment module 170 is used to adjust the charge and discharge power of the energy storage system based on the deviation information. In particular, a stable power supply is a key goal of the power system. By flexibly adjusting the charge and discharge power of the energy storage system based on the deviation information, problems such as voltage fluctuations and frequency instability caused by load and generation deviations can be avoided. For example, in the face of a large positive deviation in the short term (i.e., the actual load far exceeds the expectation), the system may quickly increase the discharge rate of the energy storage device and reduce unnecessary charging behaviors; on the contrary, if a negative deviation is detected (i.e., the actual load is lower than the expectation), it may give priority to using the excess power to charge the energy storage system to reserve energy for high-demand periods in the future. This process requires the energy storage management system to have a high degree of flexibility and response speed and be able to quickly adapt to the changing grid conditions.

[0080] In summary, the day-ahead and real-time coordinated scheduling system 100 for the electric energy substitution load and energy storage system based on the embodiments of the present application is elucidated. It first obtains the historical data of the electric energy substitution load and renewable energy power generation, then inputs them into the load prediction model and power generation prediction model respectively to obtain the day-ahead load prediction result and the day-ahead power generation prediction result, and generates a day-ahead scheduling plan based on the obtained prediction results and the preset optimization objectives. Subsequently, it obtains the actual electric energy substitution load and renewable energy power generation data, calculates the deviation information between the actual data and the prediction data according to the day-ahead scheduling plan, and finally adjusts the charge and discharge power of the energy storage system based on the deviation information. In this way, the coordinated operation efficiency of the electric energy substitution equipment cluster and renewable energy can be effectively improved, and thus the economic operation and safety stability of the system can be effectively ensured.

[0081] The basic principles of the present application have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, benefits, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, benefits, effects, etc. are essential for each embodiment of the present application. In addition, the specific details of the above application are only for the purpose of illustration and easy understanding, rather than limitations, and the above details do not limit the present application to necessarily adopt the above specific details for implementation.

Claims

1. A day-ahead and real-time collaborative scheduling system for electric energy substitution load and energy storage system, characterized in that, Including: A historical data acquisition module, configured to acquire historical data of the electricity substitution load and historical data of renewable energy power generation data; A day-ahead load forecasting module, configured to input the historical data of the electricity substitution load into a load forecasting model to obtain a day-ahead load forecasting result. Among them, the day-ahead load forecasting module is configured to: perform spatio-temporal constraint transfer analysis based on the local time series feature domain on the historical data of the electricity substitution load to obtain the day-ahead load forecasting result; A power generation forecasting module, configured to input the historical data of the renewable energy power generation data into a renewable energy power generation forecasting model to obtain a day-ahead renewable energy power generation forecasting result; A scheduling plan generation module, configured to generate a day-ahead scheduling plan based on the day-ahead load forecasting result and the day-ahead renewable energy power generation forecasting result and in combination with a preset optimization objective; An actual data acquisition module, configured to acquire actual electricity substitution load data and actual renewable energy power generation data; A deviation calculation module, configured to calculate deviation information of the actual electricity substitution load data and the actual renewable energy power generation data based on the day-ahead scheduling plan; A power adjustment module, configured to adjust the charge and discharge power of the energy storage system based on the deviation information.

2. The day-ahead and real-time collaborative scheduling system for the electric energy substitution load and energy storage system according to claim 1, wherein The day-ahead load forecasting module includes: An electricity substitution load local time series encoding unit, configured to perform local time series sequence encoding on the historical data of the electricity substitution load to obtain a set of electricity substitution load local time series correlation feature vectors; An electricity substitution load local time series message passing encoding unit, configured to perform message passing encoding based on spatio-temporal double constraints on the set of electricity substitution load local time series correlation feature vectors to obtain an electricity substitution load spatio-temporal double constraint feature vector; A day-ahead load forecasting result generation unit, configured to obtain the day-ahead load forecasting result based on the electricity substitution load spatio-temporal double constraint feature vector.

3. The day-ahead and real-time coordinated scheduling system for the electric energy substitution load and energy storage system according to claim 2, wherein The electricity substitution load local time series encoding unit includes: An electricity substitution load data slicing sub-unit, configured to perform data time series slicing on the historical data of the electricity substitution load based on a preset time scale to obtain a set of electricity substitution load local time series distributions; An electricity substitution load local time series correlation feature extraction sub-unit, configured to perform sequence encoding on each electricity substitution load local time series distribution in the set of electricity substitution load local time series distributions to obtain the set of electricity substitution load local time series correlation feature vectors.

4. The day-ahead and real-time collaborative scheduling system for the electric energy substitution load and energy storage system according to claim 3, characterized in that, The electricity substitution load local time series correlation feature extraction sub-unit is configured to: perform sequence encoding based on one-dimensional convolution on each electricity substitution load local time series distribution in the set of electricity substitution load local time series distributions to obtain the set of electricity substitution load local time series correlation feature vectors.

5. The day-ahead and real-time collaborative scheduling system for the electric energy substitution load and energy storage system according to claim 2, wherein The electricity substitution load local time series message passing encoding unit includes: A time series feature sequence encoding sub-unit, configured to input the set of electricity substitution load local time series correlation feature vectors into a sequence encoder based on a recurrent neural network to obtain a set of initial encoding vectors for sequence transmission of electricity substitution load local time series correlation features; The spatio-temporal collaborative constraint factor calculation subunit is used to calculate the power substitution load local temporal correlation feature message passing spatio-temporal collaborative constraint factor for each initial coding vector of the power substitution load local temporal correlation feature in the set of initial coding vectors of the sequence transfer power substitution load local temporal correlation feature; The power substitution load spatio-temporal aggregation subunit is used to perform message passing structure modulation aggregation on each initial coding vector of the power substitution load local temporal correlation feature based on the power substitution load local temporal correlation feature message passing spatio-temporal collaborative constraint factor to obtain the power substitution load spatio-temporal double-constraint feature vector.

6. The day-ahead and real-time collaborative scheduling system for the electric energy substitution load and energy storage system according to claim 5, characterized in that, The spatio-temporal collaborative constraint factor calculation subunit includes: The temporal confidence constraint factor calculation secondary subunit is used to calculate the power substitution load local temporal correlation feature temporal confidence constraint factor for each initial coding vector of the power substitution load local temporal correlation feature in the set of initial coding vectors of the sequence transfer power substitution load local temporal correlation feature; The spatial confidence constraint factor calculation secondary subunit is used to calculate the power substitution load local temporal correlation feature spatial confidence constraint factor for each initial coding vector of the power substitution load local temporal correlation feature in the set of initial coding vectors of the sequence transfer power substitution load local temporal correlation feature; The message passing spatio-temporal collaborative constraint factor construction secondary subunit is used to construct the power substitution load local temporal correlation feature message passing spatio-temporal collaborative constraint factor for each initial coding vector of the power substitution load local temporal correlation feature based on the power substitution load local temporal correlation feature spatial confidence constraint factor and the power substitution load local temporal correlation feature temporal confidence constraint factor of each initial coding vector of the sequence transfer power substitution load local temporal correlation feature.

7. The day-ahead and real-time collaborative scheduling system for the electric energy substitution load and energy storage system according to claim 6, characterized in that The message passing spatio-temporal collaborative constraint factor construction secondary subunit is used to: perform curvature-compensation-based spatio-temporal confidence constraint collaborative optimization on the power substitution load local temporal correlation feature spatial confidence constraint factor and the power substitution load local temporal correlation feature temporal confidence constraint factor of each initial coding vector of the power substitution load local temporal correlation feature to obtain the power substitution load local temporal correlation feature message passing spatio-temporal collaborative constraint factor for each initial coding vector of the sequence transfer power substitution load local temporal correlation feature.

8. The day-ahead and real-time collaborative scheduling system for the electric energy substitution load and energy storage system according to claim 7, wherein The power substitution load spatio-temporal aggregation subunit is used to: Perform message passing structure modulation on each initial coding vector of the power substitution load local temporal correlation feature based on the power substitution load local temporal correlation feature message passing spatio-temporal collaborative constraint factor to obtain a set of structurally modulated coding vectors of the sequence transfer power substitution load local temporal correlation feature; Calculate the position-wise sum of the set of structurally modulated coding vectors of the sequence transfer power substitution load local temporal correlation feature to obtain the power substitution load spatio-temporal double-constraint feature vector.

9. The day-ahead and real-time collaborative scheduling system for the electric energy substitution load and energy storage system according to claim 8, wherein The day-ahead load prediction result generation unit is configured to: perform feature decoding on the spatio-temporal dual-constraint feature vector of the electricity substitution load based on an RNN model to obtain the day-ahead load prediction result.

Images