Multi-energy cooperative scheduling method and system for integrated energy system based on data driving
By constructing a scheduling correlation map between the power generation and power consumption sides and a multi-energy collaborative scheduling model based on Transformer, the problem of poor adaptability of traditional scheduling schemes is solved, and precise and efficient scheduling of integrated energy systems is achieved, meeting the needs of supply and demand balance and optimization.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-20
- Publication Date
- 2026-03-27
AI Technical Summary
Traditional integrated energy system dispatching cannot fully integrate data related to multiple energy sources and lacks effective sorting of the coordination logic between power generation and consumption, resulting in dispatching schemes that are difficult to adapt to complex energy unit scenarios. The generated dispatching schemes have poor adaptability and low efficiency, making it difficult to meet the needs of precise and efficient multi-energy coordinated dispatching.
By collecting operating parameters, historical scheduling data, and strategy factor data of each energy unit within the integrated energy system, the characteristics of multi-energy coordinated scheduling are determined, a scheduling correlation map of the power generation side and the power consumption side is constructed, and the optimal scheduling scheme is determined using a Transformer-based multi-energy coordinated scheduling model.
It achieves the precision and efficiency of multi-energy coordinated scheduling of integrated energy systems, meets the system's supply and demand balance and optimization needs, and ensures the system's scheduling optimization goals.
Smart Images
Figure CN121749356A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of energy big data, and in particular to a data-driven comprehensive energy system multi-energy collaborative scheduling method and system. BACKGROUND
[0002] Comprehensive energy system multi-energy collaborative scheduling is crucial to improving energy utilization efficiency and ensuring supply and demand balance, and scheduling scheme formulation is its core requirement. Existing technologies rely on experience or a single model for scheduling, which plays a certain role in simple energy scenarios, but as the number of energy units increases and the supply and demand relationship becomes more complex, traditional methods have obvious limitations. Traditional scheduling cannot accurately determine the characteristics of multi-energy collaborative scheduling, and it is also difficult to build a correlation graph between the power generation side and the power consumption side, resulting in poor adaptability and low efficiency of the generated scheduling scheme, which is difficult to meet the precise and efficient multi-energy collaborative scheduling requirements of comprehensive energy systems. SUMMARY
[0003] The present application provides a data-driven comprehensive energy system multi-energy collaborative scheduling method and system, which solves the technical problem that traditional comprehensive energy system scheduling fails to fully integrate multi-type energy related data and lacks effective analysis of collaborative logic between power generation and power consumption, making it difficult for the scheduling scheme to adapt to complex energy unit scenarios.
[0004] In a first aspect, the present application provides a data-driven comprehensive energy system multi-energy collaborative scheduling method, which comprises: determining multi-energy collaborative scheduling characteristics according to the operating parameters, historical scheduling data and strategy factor data of each energy unit in the comprehensive energy system; performing priority measurement through the multi-energy collaborative scheduling characteristics to determine a first type of scheduling correlation graph of the energy units on the power generation side and a second type of scheduling correlation graph of the energy units on the power consumption side; and determining an optimal scheduling scheme based on the first type of scheduling correlation graph of the energy units on the power generation side and the second type of scheduling correlation graph of the energy units on the power consumption side, in combination with an energy scheduling strategy library using a multi-energy collaborative scheduling model based on Transformer.
[0005] In a second aspect, the present application provides a data-driven comprehensive energy system multi-energy collaborative scheduling system, which comprises: a multi-energy collaborative scheduling characteristic acquisition module configured to determine multi-energy collaborative scheduling characteristics according to the operating parameters, historical scheduling data and strategy factor data of each energy unit in the comprehensive energy system; a scheduling correlation graph acquisition module configured to perform priority measurement through the multi-energy collaborative scheduling characteristics to determine a first type of scheduling correlation graph of the energy units on the power generation side and a second type of scheduling correlation graph of the energy units on the power consumption side; and an optimal scheduling scheme acquisition module configured to determine an optimal scheduling scheme based on the first type of scheduling correlation graph of the energy units on the power generation side and the second type of scheduling correlation graph of the energy units on the power consumption side, in combination with an energy scheduling strategy library using a multi-energy collaborative scheduling model based on Transformer.
[0006] One or more technical solutions provided in the present application have at least the following technical effects or advantages: The present application acquires multi-energy collaborative scheduling features and a first type of scheduling correlation graph on the power generation side and a second type of scheduling correlation graph on the power consumption side through processing such as feature screening and priority measurement by collecting operation parameters, historical scheduling data and policy factor data of each energy unit in the integrated energy system, mines the spatiotemporal coupling relationship of the energy units to determine the optimal output and power consumption state, adjusts in combination with a historical optimal scheduling case library and a compensation model, and thus accurately determines the optimal scheduling scheme of the next time slice of the integrated energy system, makes the multi-energy collaborative scheduling result of the integrated energy system more accurate and efficient, meets the system supply-demand balance and optimization demand, and achieves the technical effects of accuracy and efficiency of the multi-energy collaborative scheduling of the integrated energy system and protection of the system supply-demand balance and scheduling optimization target. BRIEF DESCRIPTION OF DRAWINGS
[0007] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and all other drawings obtained by those of ordinary skill in the art without creative effort based on these drawings also belong to the protection scope of the present application.
[0008] Figure 1 is a flowchart of the data-driven integrated energy system multi-energy collaborative scheduling method provided by the embodiments of the present application.
[0009] Figure 2 is a structural schematic diagram of the data-driven integrated energy system multi-energy collaborative scheduling system provided by the embodiments of the present application.
[0010] Legend of the drawings: multi-energy collaborative scheduling feature acquisition module 1, scheduling correlation graph acquisition module 2, optimal scheduling scheme acquisition module 3. DETAILED DESCRIPTION
[0011] The present application provides a data-driven integrated energy system multi-energy collaborative scheduling method and system, and solves the technical problem that the traditional integrated energy system scheduling fails to fully integrate multiple types of energy related data, lacks effective analysis of the collaborative logic on the power generation and power consumption sides, and thus the scheduling scheme is difficult to adapt to complex energy unit scenarios.
[0012] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort belong to the protection scope of the present application.
[0013] It should be noted that the terms "first", "second", etc. in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or server including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but can include other steps or modules not clearly listed or inherent to these processes, methods, products or devices.
[0014] Embodiment one, as shown in a data-driven integrated energy system multi-energy collaborative scheduling method, wherein the method comprises: Figure 1 According to the operation parameters, historical scheduling data and policy factor data of each energy unit in the integrated energy system, the multi-energy collaborative scheduling characteristics are determined. According to the operation parameters, historical scheduling data and policy factor data of each energy unit in the integrated energy system, the multi-energy collaborative scheduling characteristics are determined.
[0015] In the embodiments of the present application, the integrated energy system is an energy system that contains various energy units such as power generation side and power consumption side, and realizes multi-energy collaborative scheduling by integrating unit operation parameters, historical scheduling data and other information.
[0016] Specifically, the specific way to determine the multi-energy collaborative scheduling characteristics is: on the one hand, real-time environmental parameters are introduced as supplementary features to enrich the feature dimension; on the other hand, based on the operation parameters, historical scheduling data and policy factor data of each energy unit in the integrated energy system, these basic data are combined with the above-mentioned supplementary features, and through feature screening operation, the multi-energy collaborative scheduling characteristics are finally determined.
[0017] The priority is measured by the multi-energy collaborative scheduling characteristics, and a first type of scheduling association graph of the power generation side energy unit and a second type of scheduling association graph of the power consumption side energy unit are determined.
[0018] Optionally, first, according to the carbon emission intensity and output stability of the output end of each energy unit, a power generation priority quantitative index is configured; then, using the power generation priority quantitative index, a priority measurement is carried out on the power generation side characteristics in the multi-energy collaborative scheduling characteristics, and a first type of scheduling association graph of the power generation side is constructed.
[0019] Then, according to the load importance, response flexibility and interruption loss of the input end of each energy unit, a power consumption priority quantitative index is configured; then, using the power consumption priority quantitative index, a second priority measurement is carried out on the power consumption side characteristics in the multi-energy collaborative scheduling characteristics, and a second type of scheduling association graph of the power consumption side is constructed.
[0020] The first scheduling association graph based on the power generation side energy unit, the second scheduling association graph of the power consumption side energy unit, and the energy scheduling strategy library are used to determine the optimal scheduling scheme by using the multi-energy collaborative scheduling model based on the Transformer.
[0021] In the embodiments of the present application, the Transformer is an encoder-decoder architecture constructed entirely based on the self-attention mechanism, can process sequence data in parallel and accurately capture long-distance dependencies, and provides a deep learning architecture for the underlying support of mainstream AI models in the fields of NLP, CV, etc.
[0022] In an embodiment of the present application, first, by the energy supply and demand mapping relationship, the power generation unit output node in the first scheduling association graph of the power generation side is associated with the power consumption unit demand node in the second scheduling association graph of the power consumption side, which is used as the input variable set of the multi-energy collaborative scheduling model. Then, the multi-energy collaborative scheduling model based on the Transformer is used to mine the spatiotemporal coupling relationship of each energy unit, and then determine the optimal output state and optimal power consumption state of each energy unit in the next time slice, and finally generate the optimal scheduling scheme.
[0023] Further, the method provided in the embodiments of the present application comprises: Real-time environmental parameters are used as supplementary features. The multi-energy collaborative scheduling features are determined by performing feature screening on the operating parameters, historical scheduling data and strategy factor data of the energy units in combination with the supplementary features.
[0024] Specifically, first, real-time environmental parameters are obtained and used as supplementary features. Real-time environmental data of the area where the integrated energy system is located is collected by using a sensing device, including parameters such as light intensity, wind speed, environmental temperature, humidity, etc. These parameters will directly affect the operating state of the energy unit, for example, the light intensity determines the actual output upper limit of the photovoltaic unit, and the wind speed affects the power generation stability of the wind power unit. By including these real-time environmental parameters in the feature system, the problem of incomplete feature dimension caused by relying only on the data of the energy unit itself in the traditional technology can be solved, and the supplementary features are formed.
[0025] Next, the supplementary features are normalized to generate standard feature values, and then the standard feature values are combined with the operating parameter feature values, historical scheduling data feature values and strategy factor data feature values corresponding to each energy unit to obtain a multi-dimensional energy feature set. Then, the variance analysis is used to sort the feature importance of the multi-dimensional energy feature set, and the multi-energy collaborative scheduling features are obtained by sorting and screening. This step is described in detail in the subsequent content.
[0026] Further, the method provided in the embodiments of the present application comprises: According to the supplementary feature, a standard feature value is generated by normalization processing, and a multi-dimensional energy feature set is obtained by combining the operation parameter feature value, the historical scheduling data feature value and the strategy factor data feature value corresponding to each energy unit.
[0027] Optionally, first, the real-time environmental supplementary features determined in the foregoing step, such as light intensity, wind speed and environmental temperature, are processed by using a Min-Max normalization method: the maximum value and the minimum value of a certain supplementary feature in historical data are calculated, and then all values of the feature are mapped to a unified interval of 0-1 through the formula (current feature value-minimum value) / (maximum value-minimum value), to generate a standard feature value eliminating the dimensional influence.
[0028] Subsequently, the standard feature value is integrated with three types of core feature values corresponding to each energy unit, including the operation parameter feature value of the energy unit, such as the actual output value of photovoltaic, the unit fuel consumption value of thermal power and the residual capacity value of energy storage; the historical scheduling data feature value, such as the matching deviation value of the power generation side output and the power consumption side demand in the same time period in the past, and the execution efficiency value of the historical scheduling scheme; and the strategy factor data feature value, such as the cost coefficient, the carbon emission coefficient and the response speed coefficient under the current scheduling strategy, to finally form a multi-dimensional energy feature set covering the energy itself attributes, external environmental influences, historical experiences and strategy orientations.
[0029] Next, the multi-dimensional energy feature set is sorted in importance by using variance analysis, and the multi-energy collaborative scheduling feature is selected. There are often redundant features in the multi-dimensional feature set, for example, the humidity data fluctuation is very small in a certain region in a certain season, and the influence on energy output can be ignored, which needs to be removed by scientific methods. The specific process is as follows: first, the multi-dimensional energy feature set is taken as the analysis object, and the contribution of the feature to the difference of the comprehensive energy scheduling result is taken as the evaluation index, and the variance analysis method is used to calculate the variance proportion of each feature. The higher the variance proportion, the more significant the influence of the feature on the scheduling result, for example, the variance proportion of the light intensity feature is high, because it directly determines the upper limit of the output of the photovoltaic unit, and then affects the overall supply and demand balance.
[0030] Subsequently, all features are sorted according to the variance proportion from high to low, and the features with key influence on scheduling decision are retained, such as light intensity, wind speed, photovoltaic output value, high-priority load power and carbon emission coefficient, and the redundant features with very low variance proportion are removed, such as humidity data with very small fluctuation and historical redundant records irrelevant to the current scheduling strategy, to finally select the multi-energy collaborative scheduling feature that can accurately support the subsequent scheduling decision.
[0031] Further, the method provided in the embodiments of the present application comprises: a power generation priority quantification index is configured based on carbon emission intensity and output stability of each energy unit output end; the power generation side feature in the multi-energy collaborative scheduling feature is measured in priority once using the power generation priority quantification index, and a first type of scheduling correlation graph is constructed.
[0032] Specifically, a power generation priority quantification index is configured based on carbon emission intensity and output stability of each energy unit output end. Weighted summation is used to realize quantification, and the specific process is as follows: first, carbon emission data of each power generation unit in the past year is collected, and the average carbon emission per kilowatt-hour of power generation is calculated as the carbon emission intensity; then, the hourly output data of each unit in the past three months is collected, and the standard deviation of the deviation between the actual output and the rated output is calculated, and the smaller the standard deviation, the higher the output stability.
[0033] Then, weights are allocated according to actual strategy requirements (such as green electricity priority or stability priority), if green electricity priority is given priority to, the carbon emission intensity weight is set to 0.6, and the output stability is set to 0.4, and vice versa; subsequently, the values of the two indexes are standardized to the interval of 0-1, the lower the carbon emission intensity and the higher the stability, the higher the score, and the weighted sum is obtained, thereby obtaining the power generation priority quantification index of each unit.
[0034] Next, the quantification index is used to measure the priority of the power generation side feature in the multi-energy collaborative scheduling feature once, and a first type of scheduling correlation graph is constructed. The specific process is as follows: the power generation side feature is extracted from the multi-energy collaborative scheduling feature, including the rated output, the maximum output and the real-time output prediction value of each unit; the power generation priority quantification index is corresponded to these features one by one, and the power generation units are sorted in descending order of index score; then, each power generation unit is taken as a graph node, and its priority, rated output and other attributes are labeled, and according to the complementary relationship of the output (such as wind power and energy storage linkage), the correlation edges are established between the related nodes, and finally a first type of scheduling correlation graph containing priority and unit characteristics is formed.
[0035] By using the weighted summation method to quantitatively determine the priority index and the sorting correlation method to construct the graph, the scheduling priority of the power generation side energy unit is clarified and the correlation relationship is visualized, thereby laying a precise foundation for subsequent matching of power supply and demand on the power consumption side.
[0036] Further, the method provided in the embodiments of the present application comprises: a power consumption priority quantification index is configured based on the importance of the load, the response flexibility and the interruption loss of the input end of each energy unit; the power consumption side feature in the multi-energy collaborative scheduling feature is measured in priority twice using the power consumption priority quantification index, and a second type of scheduling correlation graph is constructed.
[0037] Specifically, first, based on the load importance, response flexibility, and interruption loss of each energy unit input end, the electricity consumption priority quantitative indicators are configured. The weighted scoring method is used for quantification, and the specific process is as follows: first, the basic data of each power unit is collected, the load importance is divided into grades according to the influence range and degree, for example, the hospital load is set to 5 levels due to the influence on life safety, the ordinary commercial load is set to 3 levels, and the residential lighting load is set to 2 levels; the response flexibility is divided into grades according to the length and power range of the peak-shifting adjustment, for example, the industrial production load can be peak-shifting for 8 hours and is set to 4 levels, and the residential air conditioning load can be peak-shifting for 2 hours and is set to 2 levels; the interruption loss is calculated according to the direct economic loss per hour of interruption, for example, the data center has an interruption loss of 100,000 yuan per hour, and the ordinary residential load has an interruption loss of 500 yuan per hour.
[0038] Then, according to the current scheduling target, for example, the load importance weight is set to 0.5 for safety priority, and the interruption loss weight is set to 0.4 for economic priority, and the weights of the three indicators are allocated; subsequently, the load importance grade, response flexibility grade, and interruption loss value are standardized to the 0-1 interval, wherein the higher the grade and the greater the loss correspond to the higher standardized value, and the score is calculated according to the formula (load importance standardized value x weight 1) + (response flexibility standardized value x weight 2) + (interruption loss standardized value x weight 3), which is the electricity consumption priority quantitative indicator of each power unit.
[0039] After that, using the electricity consumption priority quantitative indicator, the electricity consumption side characteristics in the multi-energy collaborative scheduling characteristics are measured again, and a two-class scheduling correlation graph is constructed. The specific process is as follows: the electricity consumption side core characteristics are extracted from the multi-energy collaborative scheduling characteristics, including the rated electricity consumption power, real-time electricity demand, and peak-shifting time window of each power unit; the electricity consumption priority quantitative indicator of each power unit is bound to the corresponding electricity consumption side characteristics, and all power units are prioritized according to the quantitative indicator score from high to low, for example, the hospital load has a score of 0.92 and ranks first, the data center load has a score of 0.85 and ranks second, and the ordinary industrial load has a score of 0.6 and ranks third.
[0040] Finally, taking each power unit as an independent node, the priority ranking, rated electricity consumption power, and peak-shifting time window of the node are marked in the graph, and according to the power supply line correlation relationship of the power unit, such as the multiple shop loads in the same commercial building sharing a power supply line, and the regional location correlation relationship, such as the industrial loads in the same industrial park, the correlation edges are established between the related nodes, and finally the two-class scheduling correlation graph containing the priority, core characteristics, and correlation relationship of the power unit is formed.
[0041] The quantification calculation of the electricity priority is realized by the weighted scoring method, the graph is constructed by the node association method, the accurate definition of the electricity side energy unit scheduling priority and the visualization of the correlation are achieved, and a clear electricity demand basis is provided for the subsequent supply-demand matching of the generation side one type of scheduling correlation graph.
[0042] Further, the method provided by the embodiment of the application comprises: By the energy supply and demand mapping relationship, the power generation unit output node in the one type of scheduling correlation graph is associated and aligned with the electricity consumption unit demand node in the two types of scheduling correlation graph, as an input variable set of the multi-energy collaborative scheduling model; the time-space coupling relationship of each energy unit is mined by using the multi-energy collaborative scheduling model based on the Transformer, the optimal output state and the optimal electricity consumption state of each energy unit in the next time slice are determined, and the optimal scheduling scheme is generated.
[0043] Specifically, first, the association and alignment of the power generation and electricity consumption nodes are realized by the energy supply and demand mapping relationship. The specific process is as follows: first, the core attributes of each power generation unit output node are extracted from the one type of scheduling correlation graph, including the power generation priority, that is, the quantization result based on the carbon emission intensity and the output stability in the foregoing step; the rated output range, that is, the maximum power generation value and the minimum power generation value; the real-time output prediction value, combined with the real-time environment parameter calculation in the foregoing step, such as the photovoltaic output prediction according to the light. Then, the core attributes of each electricity consumption unit demand node are extracted from the two types of scheduling correlation graph, including the electricity consumption priority, that is, the quantization result based on the load importance, the response flexibility and the interruption loss in the foregoing step; the rated electricity consumption range, that is, the maximum consumption value and the minimum consumption value; the real-time electricity consumption demand value.
[0044] Subsequently, combined with the current strategy in the energy scheduling strategy library, for example, the green electricity priority is matched with the wind power / photovoltaic node, the economic priority is matched with the low-cost thermal power node, according to the rule that the high-priority electricity consumption node is preferentially associated with the high-priority power generation node and the real-time output prediction value of the power generation node covers the real-time demand value of the electricity consumption node, the power generation unit output node is corresponded to the electricity consumption unit demand node one by one, for example, the hospital electricity consumption node (high electricity consumption priority) is associated with the thermal power / energy storage power generation node (high power generation priority, output stability), and the ordinary industrial electricity consumption node (low electricity consumption priority) is associated with the wind power node (output fluctuation acceptable). Finally, the combination of the aligned power generation node attributes-electricity consumption node attributes, together with the current strategy factors, such as the green electricity proportion requirement and the cost control threshold, are used as the input variable set of the multi-energy collaborative scheduling model.
[0045] Then a Transformer-based multi-energy collaborative scheduling model is constructed and applied. The model construction process is as follows: a Transformer model with an encoder-decoder structure is adopted, the encoder part is provided with 6 encoder layers, each layer contains a multi-head self-attention mechanism for capturing the spatial correlation between different energy units, such as the complementary relationship between photovoltaic and energy storage, and a feedforward neural network for processing time sequence features, such as the output change of different time slices; the decoder part is provided with 6 decoder layers, which are connected to the encoder output and the target sequence through a cross-attention mechanism, and the scheduling state of the next time slice, to ensure that the output meets the supply and demand requirements.
[0046] The model training process: the historical optimal scheduling case library mentioned above is used as the training sample, each sample contains three parts, the input data, the input variable set of a certain period, the real-time environmental parameters of the period, and the energy unit operation data of the previous three time slices; the label data, the optimal output state of the period verified by practice, such as the actual optimal output value of each power generation unit; the optimal power consumption state, such as the actual optimal power consumption value of each power consumption unit. During training, the mean square error is used as the loss function, the difference between the predicted output / power consumption state of the model and the label data is calculated, and the attention weight and network parameters of the model are adjusted through back propagation iteration. The model performance is verified once every 100 iterations, the prediction error is calculated using the verification set sample, and the training of the model is completed until the mean square error of the verification set is less than the set threshold.
[0047] The model application process: the current input variable set, the current real-time environmental parameters, and the energy unit operation data of the previous three time slices obtained in the above steps are input into the trained model, the encoder extracts the spatial correlation of each power generation / power consumption unit through the multi-head self-attention mechanism, such as the energy complementary relationship between wind power and the power grid, and captures the time sequence change through the feedforward neural network, such as the decay trend of photovoltaic output over time. The decoder generates the optimal output state of each power generation unit and the optimal power consumption state of each power consumption unit in the next time slice based on the encoder output and the cross-attention mechanism, and finally integrates these states to generate the optimal scheduling scheme.
[0048] Through attribute matching method, the precise association and alignment of power generation and power consumption nodes are realized, the Transformer-based multi-energy collaborative scheduling model is constructed and trained to mine the spatio-temporal coupling relationship, and the optimal scheduling scheme with strong foresight and accurate supply-demand matching for the integrated energy system is achieved.
[0049] Further, the method provided by the embodiments of the present application comprises: According to the nodes of the first type of scheduling association graph and the nodes of the second type of scheduling association graph, energy scheduling semantic features are extracted; based on a historical optimal scheduling case library, the energy supply and demand mapping relationship is established in combination with the energy scheduling semantic features, and a compensation model is constructed through a machine learning incremental algorithm to adjust the optimal scheduling scheme in real time.
[0050] In one embodiment, first, energy scheduling semantic features are extracted according to the nodes of the first type of scheduling association graph and the nodes of the second type of scheduling association graph. Feature encoding is used to extract the core attributes of each node from the power generation nodes of the first type of scheduling association graph, such as power generation priority, rated output range, carbon emission intensity, and output stability; and to extract the core attributes of each node from the power consumption nodes of the second type of scheduling association graph, such as power consumption priority, rated power consumption range, load importance, response flexibility, and interruption loss.
[0051] Then, these attributes are converted into quantifiable numerical vectors according to predetermined rules, for example, high power generation priority corresponds to a numerical value of 1, medium corresponds to 0.5, and low corresponds to 0.2; the output stability is normalized according to the standard deviation and takes a numerical value in the range of 0-1; all attribute numerical vectors of the same node are spliced to form the energy scheduling semantic features of the node; and finally, the semantic features of all power generation and power consumption nodes are integrated to obtain a system-level energy scheduling semantic feature set.
[0052] Then, based on the historical optimal scheduling case library and in combination with the semantic features, an energy supply and demand mapping relationship is established, and a compensation model is constructed through a machine learning incremental algorithm. When establishing the energy supply and demand mapping relationship, a case matching method is used, and the specific process is as follows: from the historical optimal scheduling case library, the energy scheduling semantic features (consistent with the current system semantic features) and the corresponding supply and demand mapping relationship of each historical case are extracted, for example, the corresponding rule of the high-priority hospital power consumption node associated with the high-stability thermal power node in a certain historical case.
[0053] Then, the cosine similarity of the current system semantic features and the semantic features of each historical case is calculated, and the top 5 historical cases with the highest similarity are selected; the supply and demand mapping relationship of these 5 cases is statistically analyzed, and the mapping rule with the highest frequency is retained, for example, the high-priority power consumption node is associated with the high-priority power generation node; and in combination with the current scheduling strategy, such as supplementing the high-priority power consumption node associated with the wind power / photovoltaic node when green electricity is preferred, the energy supply and demand mapping relationship that meets the current system demand is finally established.
[0054] In constructing the compensation model, the random forest incremental algorithm is selected, and the specific process is as follows: in the model construction stage, the historical scheduling data is used as the basic data set, the samples in the data set include three parts, the input features are the energy scheduling semantic features in a certain period of history, the output / prediction value of the optimal scheduling scheme, and the actual output / prediction value of the period, the bias value is the difference between the actual running value and the prediction value, the adjustment label is the effective adjustment measure verified by artificial verification or practical test, for example, when the actual output of photovoltaic is 10MW lower than the prediction, the scheduling energy storage discharges 8MW+power grid energy supplement 2MW, etc.
[0055] Then the random forest model is initialized, 100 decision trees are set as the basic model, the model is trained with the basic data set, the mapping relationship between the input features and the adjustment label is established through the decision tree splitting rule, such as taking the output bias value>5MW as the splitting node, and the basic model construction is completed. In the model training stage, the incremental learning mechanism is adopted, when new scheduling data (such as the actual running bias and adjustment measures added in the day) is generated, it is not necessary to retrain all decision trees, only a small number of decision trees are added in the existing random forest, 10 trees are added each time, the new data is used to train the added decision trees, and then the original model is integrated, at the same time, the old decision trees with an accuracy lower than the threshold value of 80% are removed from the model, so that the model can adapt to the changes of system operation in real time.
[0056] In the model application, the input is the energy scheduling semantic features of the current system, the prediction output / power value of the optimal scheduling scheme, and the real-time collected actual output / power value, the model calculates the bias value, matches the corresponding adjustment label, and outputs the specific adjustment amount, for example, the thermal power output needs to be increased by 5MW, and the certain industrial power unit needs to be peak shaving by 3MW, so as to realize real-time adjustment of the optimal scheduling scheme.
[0057] Through the feature coding method, the energy scheduling semantic features are extracted, the case matching method is used to establish the energy supply and demand mapping relationship, and the random forest incremental algorithm is used to construct the compensation model, so that the dynamic correction of the optimal scheduling scheme of the comprehensive energy system is realized, the adaptability of the scheduling scheme to the actual operation demand is improved, and the stability and accuracy of the system scheduling are enhanced.
[0058] In summary, the comprehensive energy system multi-energy collaborative scheduling method based on data driving provided by the embodiment has the following technical effects: This application collects operating parameters, historical dispatch data, and strategy factor data of each energy unit within an integrated energy system. Through feature filtering and priority measurement, it obtains multi-energy coordinated dispatch characteristics and a type-one dispatch correlation map on the generation side and a type-two dispatch correlation map on the consumption side. It calculates the spatiotemporal coupling relationship of each energy unit and the optimal output and consumption status for the next time slice. Adjustments are made using an energy dispatch strategy library, a historical best dispatch case library, and a compensation model, thereby accurately determining the optimal dispatch scheme for the integrated energy system. This makes the multi-energy coordinated dispatch results of the integrated energy system more precise and efficient, meeting the system's supply and demand balance and optimization requirements. It achieves the technical effect of ensuring the accuracy and efficiency of multi-energy coordinated dispatch in the integrated energy system, guaranteeing the system's supply and demand balance and dispatch optimization goals.
[0059] Example 2, as Figure 2 As shown, based on the same inventive concept as in Embodiment 1 above, this application provides a data-driven integrated energy system multi-energy coordinated scheduling system, the system comprising: Multi-energy coordinated scheduling feature acquisition module 1 is used to determine multi-energy coordinated scheduling features based on the operating parameters, historical scheduling data and strategy factor data of each energy unit in the integrated energy system.
[0060] The scheduling association graph acquisition module 2 is used to perform priority measurement through the multi-energy coordinated scheduling features to determine the first type of scheduling association graph of the power generation side energy unit and the second type of scheduling association graph of the power consumption side energy unit.
[0061] The optimal scheduling scheme acquisition module 3 determines the optimal scheduling scheme based on the first type of scheduling association map of the power generation side energy unit and the second type of scheduling association map of the power consumption side energy unit, combined with the energy scheduling strategy library and the Transformer-based multi-energy collaborative scheduling model.
[0062] Furthermore, the multi-energy cooperative scheduling feature acquisition module 1 is used to perform the following steps: Real-time environmental parameters are used as supplementary features; based on the operating parameters, historical scheduling data and strategy factor data of each energy unit, feature screening is performed in conjunction with the supplementary features to determine multi-energy coordinated scheduling features.
[0063] Furthermore, the multi-energy cooperative scheduling feature acquisition module 1 is used to perform the following steps: The standard characteristic values generated by normalizing according to the supplementary features are combined with the operation parameter characteristic values, the historical scheduling data characteristic values and the strategy factor data characteristic values corresponding to each energy unit to obtain a multi-dimensional energy feature set; variance analysis is used to sort the feature importance of the multi-dimensional energy feature set to filter the multi-energy collaborative scheduling features.
[0064] Further, the scheduling correlation atlas acquisition module 2 is configured to perform the following steps: Based on the carbon emission intensity and output stability of the output end of each energy unit, a power generation priority quantitative index is configured; the power generation side features in the multi-energy collaborative scheduling features are measured in priority once using the power generation priority quantitative index to construct a first type of scheduling correlation atlas.
[0065] Further, the scheduling correlation atlas acquisition module 2 is configured to perform the following steps: Based on the load importance, response flexibility and interruption loss of the input end of each energy unit, a power consumption priority quantitative index is configured; the power consumption side features in the multi-energy collaborative scheduling features are measured in priority twice using the power consumption priority quantitative index to construct a second type of scheduling correlation atlas.
[0066] Further, the optimal scheduling scheme acquisition module 3 is configured to perform the following steps: The power generation unit output nodes in the first type of scheduling correlation atlas and the power consumption unit demand nodes in the second type of scheduling correlation atlas are associated and aligned as an input variable set of the multi-energy collaborative scheduling model through an energy supply and demand mapping relationship; a multi-energy collaborative scheduling model based on Transformer is used to mine the spatiotemporal coupling relationship of each energy unit to determine the optimal output state and the optimal power consumption state of each energy unit in the next time slice to generate the optimal scheduling scheme.
[0067] Further, the optimal scheduling scheme acquisition module 3 is configured to perform the following steps: Energy scheduling semantic features are extracted according to the nodes of the first type of scheduling correlation atlas and the nodes of the second type of scheduling correlation atlas; an energy supply and demand mapping relationship is established based on a historical optimal scheduling case library in combination with the energy scheduling semantic features, and a compensation model is constructed through a machine learning incremental algorithm to adjust the optimal scheduling scheme in real time.
[0068] The data-driven comprehensive energy system multi-energy collaborative scheduling system provided in the embodiments of the present application can perform the data-driven comprehensive energy system multi-energy collaborative scheduling method provided in any embodiment of the present application, and has the corresponding functional modules and beneficial effects of the execution method.
[0069] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, however, any number of different modules can be used and run on the user terminal and / or server, the various units and modules are only divided according to the functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific name of each functional unit is only for the convenience of mutual differentiation, and does not serve to limit the protection scope of the present application.
[0070] The above detailed description does not constitute a limitation on the protection scope of the present application. Those skilled in the art should understand that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application shall be included in the protection scope of the present application. In some cases, the actions or steps described in the present application can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or can be advantageous.
Claims
1. A data-driven integrated energy system multi-energy collaborative scheduling method, characterized in that, The method comprises: According to the operation parameters, historical scheduling data and policy factor data of each energy unit in the integrated energy system, the multi-energy collaborative scheduling characteristics are determined; Through the priority measurement of the multi-energy collaborative scheduling characteristics, a first type of scheduling correlation graph of the power generation side energy unit and a second type of scheduling correlation graph of the power consumption side energy unit are determined; Based on the first type of scheduling correlation graph of the power generation side energy unit and the second type of scheduling correlation graph of the power consumption side energy unit, and combined with the energy scheduling strategy library, a multi-energy collaborative scheduling model based on Transformer is used to determine the optimal scheduling scheme.
2. The method of claim 1, wherein, According to the operation parameters, historical scheduling data and policy factor data of each energy unit in the integrated energy system, the multi-energy collaborative scheduling characteristics are determined, and the method comprises: Taking real-time environmental parameters as supplementary features; According to the operation parameters, historical scheduling data and policy factor data of each energy unit, and combined with the supplementary features, the multi-energy collaborative scheduling characteristics are determined through feature screening.
3. The method of claim 2, wherein, Combined with the supplementary features, the multi-energy collaborative scheduling characteristics are determined through feature screening, and the method comprises: According to the supplementary features, standard feature values are generated through normalization processing, and the standard feature values are combined with the operation parameter feature values, historical scheduling data feature values and policy factor data feature values corresponding to each energy unit to obtain a multi-dimensional energy feature set; The multi-dimensional energy feature set is sorted in terms of feature importance using variance analysis, and the multi-energy collaborative scheduling characteristics are screened.
4. The method of claim 1, wherein, Through the priority measurement of the multi-energy collaborative scheduling characteristics, a first type of scheduling correlation graph of the power generation side energy unit is determined, and the method comprises: Based on the carbon emission intensity and output stability of the output end of each energy unit, a power generation priority quantitative index is configured; Using the power generation priority quantitative index, a first priority measurement is performed on the power generation side features in the multi-energy collaborative scheduling characteristics, and a first type of scheduling correlation graph is constructed.
5. The method of claim 4, wherein, Through the priority measurement of the multi-energy collaborative scheduling characteristics, a second type of scheduling correlation graph of the power consumption side energy unit is determined, and the method comprises: Based on the load importance, response flexibility and interruption loss of the input end of each energy unit, a power consumption priority quantitative index is configured; Using the power consumption priority quantitative index, a second priority measurement is performed on the power consumption side features in the multi-energy collaborative scheduling characteristics, and a second type of scheduling correlation graph is constructed.
6. The method of claim 1, wherein, Based on the first type of scheduling correlation graph of the power generation side energy unit and the second type of scheduling correlation graph of the power consumption side energy unit, and combined with the energy scheduling strategy library, a multi-energy collaborative scheduling model based on Transformer is used to determine the optimal scheduling scheme, and the method comprises: Through the energy supply and demand mapping relationship, the power generation unit output node in the first type of scheduling correlation graph is associated and aligned with the power consumption unit demand node in the second type of scheduling correlation graph, serving as an input variable set of the multi-energy collaborative scheduling model; Using the multi-energy collaborative scheduling model based on Transformer, the spatio-temporal coupling relationship of each energy unit is mined, the optimal output state and optimal power consumption state of each energy unit in the next time slice are determined, and the optimal scheduling scheme is generated.
7. The method of claim 6, wherein, The method further comprises: According to the nodes of the first type of scheduling association graph and the nodes of the second type of scheduling association graph, an energy scheduling semantic feature is extracted; Based on the historical optimal scheduling case library, combined with the energy scheduling semantic feature, an energy supply and demand mapping relationship is established, and a compensation model is constructed through a machine learning incremental algorithm to adjust the optimal scheduling scheme in real time.
8. A data-driven integrated energy system multi-energy collaborative scheduling system, characterized in that, The system comprises: A multi-energy collaborative scheduling feature acquisition module is configured to determine multi-energy collaborative scheduling features according to operating parameters, historical scheduling data, and strategy factor data of each energy unit in the integrated energy system; A scheduling association graph acquisition module is configured to perform priority measurement through the multi-energy collaborative scheduling features to determine a first type of scheduling association graph of a power generation side energy unit and a second type of scheduling association graph of a power consumption side energy unit; An optimal scheduling scheme acquisition module is configured to determine an optimal scheduling scheme by using a multi-energy collaborative scheduling model based on the first type of scheduling association graph of the power generation side energy unit and the second type of scheduling association graph of the power consumption side energy unit, and in combination with an energy scheduling strategy library.