CN-122026301-A - Urban photovoltaic heat collection and energy storage integrated energy supply method for sloping fields

CN122026301ACN 122026301 ACN122026301 ACN 122026301ACN-122026301-A

Abstract

The invention provides a photovoltaic heat collection and energy storage integrated energy supply method for a sloping field city, belongs to the technical field of renewable energy utilization, and aims to solve the problem of difficulty in cooperative scheduling of electric energy and heat energy due to response characteristic difference in the sloping field city. According to the method, a cooperative model of the electric power dispatching layer and the thermal dispatching layer is constructed, a storage battery charge-discharge instruction is optimized on a minute scale, a heat storage unit heat storage and release instruction is optimized on an hour scale, and the cooperative dispatching of heterogeneous energy sources is realized through a bidirectional interaction mechanism of converting waste light power into a heating task and feeding back a thermal power consumption plan into an electric power prediction load. The method is suitable for the optimized operation of the energy supply system of the urban building group on the sloping field.

Inventors

HUANG WENPING
LIU LANJIE
DUAN HUI
CHEN TIANJIAO
YANG FAN

Assignees

中国建筑第二工程局有限公司
中建二局(四川)建设发展有限公司
中建二局重庆建设发展有限公司
中建二局重庆实业有限公司

Dates

Publication Date: 20260512
Application Date: 20251203

Claims (10)

1. The method for supplying the integrated photovoltaic heat collection and energy storage energy to the urban sloping field is characterized by comprising the following steps of: Constructing a cooperative scheduling model, wherein the model comprises a power scheduling layer and a thermal scheduling layer, and establishing energy interaction logic between the two layers; In the power dispatching layer, taking a minute level as a first time scale, based on ultra-short-term irradiance forecast data, performing rolling optimization and generating a first instruction sequence for controlling the charge and discharge power of a storage battery, wherein the optimization takes real-time balance of power supply and demand as a primary constraint; In the thermal scheduling layer, taking an hour level as a second time scale, based on short-term weather and thermal load forecast data, rolling and optimizing and generating a second instruction sequence for controlling the heat storage power of the heat storage unit, wherein the optimizing takes dynamic balance of heat energy supply and demand as a primary constraint; executing the current time instruction in the first instruction sequence and the second instruction sequence; the energy interaction logic comprises the steps of setting the optimized and determined light rejection power in the power dispatching layer as target power of electric heating equipment in the thermal dispatching layer, and submitting a future power utilization plan of the electric heating equipment in the thermal dispatching layer to the power dispatching layer as a predicted load to participate in optimization.
2. The method of claim 1, wherein the step of performing a rolling optimization at the power scheduling layer and the thermal scheduling layer and generating the sequence of instructions further comprises: In each rolling optimization period, introducing a confidence interval of a prediction error to carry out relaxation treatment on the optimization constraint condition; The method comprises the steps of loosening the power supply and demand real-time balance constraint into a limited fluctuation range determined by a confidence interval of the ultra-short-term irradiance forecast data, and loosening the heat energy supply and demand dynamic balance constraint into a limited fluctuation range determined by a confidence interval of the short-term heat load forecast data; And performing rolling optimization based on the relaxed constraint condition to generate the first instruction sequence and the second instruction sequence.
3. The method of claim 1, wherein the step of performing a rolling optimization at the thermal schedule layer on an hour scale for a second time scale comprises: decomposing the global optimization problem of the thermal scheduling layer into a plurality of local sub-optimization problems based on the slope elevation partition or the building function partition; solving each local sub-optimization problem in parallel to generate a local heat storage and release instruction sequence corresponding to each partition; And carrying out system level coordination based on the local instruction sequences of the partitions to generate a final second instruction sequence.
4. The method of claim 1, wherein the initializing step is performed when the system is restarted after a first start-up or a long shutdown: matching the operation records of similar histories according to the current date, the real-time meteorological data and the system history operation database; taking the contemporaneous energy storage unit state data of the historical similar days as an initial state value of a current rolling optimization model; And taking the contemporaneous optimized instruction sequence of the historical similar days as an initial instruction sequence of the current rolling optimization process.
5. The method of claim 1, further comprising the periodically performed model adaptation step of: Continuously monitoring and recording the actual power generation of the photovoltaic module, the actual heat collection power of the photo-thermal module and the actual energy storage efficiency of each energy storage unit; Comparing the actual data with the predicted data output by a corresponding predicted model in the current scheduling model, and calculating model prediction deviation; When the model prediction deviation continuously exceeds a preset threshold value, triggering a model updating flow, and re-identifying and correcting parameters of the prediction model by utilizing recent historical operation data.
6. The method of claim 2, wherein the confidence interval is dynamically adaptive; The step of relaxing the optimization constraint condition specifically comprises the following steps: dynamically adjusting the width of the confidence interval according to the length of the predicted time span, wherein for a power scheduling layer, the width of the confidence interval of the ultra-short-term irradiance forecast data is increased along with the extension of the predicted time span from the current moment to the future; And/or the number of the groups of groups, And dynamically adjusting the width of the confidence interval according to the statistical characteristics of the recent prediction errors, wherein the variance of the prediction errors is calculated based on the comparison of the historical prediction data and the actual data in the rolling time window, and the width of the confidence interval is scaled according to the variance.
7. The method of claim 3, wherein the step of performing system level coordination is implemented using a distributed coordination mechanism based on neighborhood interactions, and specifically comprises: The system level coordinator calculates and transmits an initial heat storage and release power reference value to each partition according to the global heat load demand and the system supply capacity; each partition exchanges current heat storage instruction state information with the adjacent partition connected in a communication mode based on the reference value; Each partition iteratively updates the heat storage and release instruction of the partition according to the operation constraint of the partition, the instruction state of the adjacent partition and the reference value; The iterative updating process is continuously carried out until the instruction variation of two adjacent iterations of all the subareas is smaller than a preset convergence threshold value; And when the convergence condition is met, the system level coordinator gathers the final instructions of the partitions and generates the second instruction sequence.
8. The method of claim 4, wherein the step of matching historical similarity days further comprises: Constructing a comprehensive feature vector containing date type, real-time meteorological data, future meteorological forecast and known event information; Calculating the similarity between the comprehensive feature vector at the current starting moment and all recorded comprehensive feature vectors in the historical database; selecting the first K history days with highest similarity as a candidate set; And generating a set of initial state values and an initial instruction sequence through weighted average based on the historical contemporaneous state data and the instruction sequence corresponding to the candidate set, wherein the weight is positively correlated with the similarity of each candidate day.
9. The method of claim 5, wherein the model adaptation step further comprises: during the running process of the system, the current running condition of the system is identified in real time, and the running condition is divided at least according to the ambient temperature interval and the sky cloud amount grade; Establishing and maintaining an independent predictor model for each of the operating conditions; The step of continuously monitoring and comparing is carried out based on an independent predictor model corresponding to the current operation condition; The trigger model updating flow is the parameter correction for the independent predictor model corresponding to the current operation condition.
10. The method of claim 9, wherein the step of creating and maintaining a separate predictor model for each operating condition is implemented as: For each operating condition, maintaining a corresponding independent predictor model; when prediction is carried out, calculating the weight of each independent predictor model according to the matching degree of the current operation working condition and each preset working condition characteristic; weighting and fusing the output results of the independent predictor models to obtain final prediction data; The independent predictor model with higher matching degree with the current operation condition has larger weight in fusion.

Description

Urban photovoltaic heat collection and energy storage integrated energy supply method for sloping fields Technical Field The invention belongs to the technical field of renewable energy and urban energy systems, and particularly relates to a sloping field urban photovoltaic heat collection and energy storage integrated energy supply method. Background In the field of renewable energy source utilization in hillside cities, the existing energy source supply method faces a plurality of technical challenges in realizing the cooperative operation of photovoltaic, photo-thermal and energy storage systems. Due to the height difference change of the hillside land topography and the three-dimensional characteristics of building distribution, the traditional energy scheduling method is difficult to effectively coordinate energy forms with different characteristics, and the overall efficiency of system operation is directly affected. In particular, the existing method has difficulty in coordination when processing energy sources with significant differences between two physical properties of electric energy and heat energy. The electric energy has the characteristics of quick instantaneous change and high storage cost, and the heat energy has the characteristics of large transmission delay and strong system inertia. This mismatch in dynamic response characteristics makes it difficult for the system to meet both the fast regulation requirements of the grid and the steady supply requirements of the thermal load. When photovoltaic power generation is subjected to power fluctuation due to cloud cover, a power system needs to respond quickly to maintain stability, but a thermodynamic system cannot provide effective support in time due to self thermal inertia, and the difference of response speeds makes the overall operation efficiency of the system limited. The root cause of this problem is that there are inherent differences in the physical characteristics of the different energy forms, and conventional control methods fail to establish an effective cross-energy coordination mechanism. In terms of system optimization, the existing method faces contradiction between computational complexity and real-time requirements. The urban energy system of the sloping field has wide coverage and complex structure, and the global optimization of the urban energy system of the sloping field needs to process a large-scale mathematical model, so that the calculated amount is obviously increased. In practical application, the time required for optimizing and solving may exceed the dynamic response time limit of the system, so that the optimization result is difficult to meet the real-time scheduling requirement. While the use of distributed computing may reduce the local computational burden, simple task decomposition may introduce new problems, such as the optimization objective of each subsystem may conflict with a global optimization objective, or where multiple subsystems compete for limited resources at the same time. The existence of these problems complicates the coordination mechanism design for distributed optimization. Prediction uncertainty is another significant technical difficulty. The optimal scheduling of the energy system depends on the accurate prediction of variables such as photovoltaic output, load demand and the like, but the prediction precision is greatly limited due to complex microclimate phenomenon and changeable user energy consumption behaviors in the sloping field city. The prediction model based on fixed parameters is difficult to adapt to the change of system characteristics in long-term operation, and prediction errors can be amplified through the optimization link, so that the reliability of scheduling decisions is finally affected. While system robustness may be enhanced by setting a larger safety margin, this tends to result in a decrease in system operating economics. In the aspect of system initialization, the existing method is not perfect in treating the cold start problem. Reasonable setting of the initial state of the system has an important impact on the subsequent optimization effect, but in the absence of historical operational data references, how to determine the appropriate initial state value is a challenging problem. An improper initial value may result in a system that takes a longer time to enter a steady state operation or may result in unnecessary energy loss during start-up. In addition, performance maintenance in long-term operation of the system is also difficult. The system component performance may gradually decay over time, and the user energy usage pattern may also change, all of which may cause the fixed model-based scheduling strategy to gradually deviate from optimal. Because of the complexity of the system operating environment, building a model adaptation mechanism that can adapt to long-term changes faces significant challenges, particularly in maintaining model accuracy wh