Reservoir scheduling rule extraction method based on rolling simulation strategy

Abstract

The invention provides a reservoir scheduling rule extraction method based on a rolling simulation strategy, which comprises the following steps: an initial water level Z0 of a reservoir scheduling period is given, and the scheduling period t is equal to 0; the initial water level Zt and the reservoir inflow It are used as characteristic factors, the final water level Zt + 1 is used as a target value, a model Mt is trained, and Mt represents the scheduling rule of the reservoir in the tth scheduling time period; inputting the initial water level Zt and the reservoir flow It into the model Mt to obtain a decision water level Kt + 1; judging whether the decision-making water level Kt + 1 violates the reservoir dispatching constraint or not, and if yes, correcting the decision-making water level Kt + 1; setting the decision water level Kt + 1 as the initial water level Zt + 1 of the next scheduling time period, and letting t = t + 1; and judging whether the scheduling time period t meets an end condition, if so, outputting a final scheduling rule, and if not, circularly executing the steps 2-6. The practicability of the scheduling rule and the power generation benefit of the reservoir system are improved.

Description

Technical Field

[0001] The invention relates to the technical field of reservoir dispatching, and in particular to a method for extracting reservoir dispatching rules based on a rolling simulation strategy. Background Art

[0002] Reservoir dispatching is to maximize the economic and social benefits of the reservoir system by rationally arranging the storage and release of water in the reservoir under the premise of ensuring flood control safety. As the core tool to guide the operation of the reservoir system, reservoir dispatching rules not only affect the planning and construction of water conservancy projects, but also are the key factors to ensure the comprehensive benefits of the reservoir operation and management period. With the development of social economy and the impact of environmental changes, reservoir dispatching faces more complex challenges, such as increased hydrological uncertainty caused by climate change and increased demand for ecological and environmental protection.

[0003] Traditional reservoir operation methods mostly rely on empirical formulas and expert knowledge. Although this method can meet the needs of reservoir operation to a certain extent, it is difficult to adapt to complex and changing hydrological conditions due to the lack of in-depth analysis of historical data and consideration of future uncertainties. In recent years, with the development of information technology, especially data mining technology, data-driven methods have begun to be widely used in the extraction of reservoir operation rules. These methods include but are not limited to artificial neural networks (ANN), support vector machines (SVM), genetic algorithms (GA), etc. These advanced data processing technologies can mine potential patterns and laws from a large amount of historical data, providing a new way to formulate more scientific and reasonable operation strategies.

[0004] Although some progress has been made in the extraction of reservoir dispatching rules based on data mining, there are still some shortcomings. For example, most existing methods tend to ignore the problem of error accumulation in the decision-making process when constructing dispatching models, which may cause the model to gradually deviate from the actual operating state in the long-term operation, thereby affecting the reliability and accuracy of the dispatching rules. In addition, the existing methods have limited ability to respond to extreme weather events. When encountering rare high or low flow conditions, the performance of the model is often greatly reduced. These problems limit the effectiveness of existing methods in practical applications, and there is an urgent need to find new solutions. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a reservoir dispatching rule extraction method based on a rolling simulation strategy. In order to achieve the above-mentioned purpose, the present invention adopts the following technical solutions:

[0006] A reservoir dispatching rule extraction method based on a rolling simulation strategy comprises the following steps:

[0007] Step 1: Given the initial water level Z0 of the reservoir during the dispatch period, the dispatch period t = 0;

[0008] Step 2, with initial water level Z t and inflow flow I t As the characteristic factor, the final water level Z t+1 is the target value, training model M t , where M t represents the dispatching rule of the reservoir in the tth dispatching period;

[0009] Step 3: Set the initial water level Z t and inflow flow I t Input model M t , get the decision water level K t+1 ;

[0010] Step 4: Determine the decision water level K t+1 Whether the reservoir operation constraints are violated, and if so, correct them;

[0011] Step 5: Set the decision water level K t+1 Set to the initial water level Z for the next scheduling period t+1 , and let t = t + 1;

[0012] Step 6: Determine whether the scheduling period t meets the end condition. If so, output the final scheduling rule. Otherwise, loop through steps 2 to 6.

[0013] Furthermore, in step 3, the water level K is determined t+1 It is model M t The predicted final water level is different from the final water level Z obtained by optimal scheduling. t+1 There is a certain deviation value.

[0014] Furthermore, in step 4, the specific method of correction is that if the decision water level exceeds the upper limit value, it is forced to be equal to the upper limit value; if the decision water level exceeds the lower limit value, it is forced to be equal to the lower limit value.

[0015] Furthermore, in steps 2 and 3, the models used include long short-term memory neural network model, artificial neural network model, light gradient boosting machine model, and linear regression model.

[0016] Compared with the prior art, the present invention has the following beneficial effects:

[0017] 1. This invention proposes a rolling simulation strategy with decision error correction capability. By introducing a decision error correction mechanism, this method can correct the model prediction results after each scheduling cycle, effectively reducing the accumulation of errors in the decision-making process. This improvement makes the scheduling rules closer to the actual operating conditions, improves the robustness and adaptability of the model, and maintains a high scheduling accuracy even in the face of uncertain hydrological conditions in the future.

[0018] 2. The present invention uses advanced machine learning algorithms, such as long short-term memory networks (LSTM), combined with rolling simulation strategies, to automatically learn and extract effective dispatching rules from a large amount of historical data. Compared with traditional methods, the present invention not only simplifies the process of rule extraction, but also improves the practicality of the rules. Specifically, the rules extracted by the present invention can better balance the relationship between power generation benefits and environmental protection, providing technical support for achieving sustainable development.

[0019] 3. By reducing the impact of decision error accumulation, the dispatching rules extracted by the present invention can more accurately reflect the operating characteristics of the reservoir system, thereby helping to improve the power generation efficiency of the reservoir system. The experimental results show that the LSTM-RS model proposed in the present invention performs well in terms of power generation, with a difference of only 80 million kWh compared with the optimization scheme. Compared with other models, the LSTM-RS model has increased power generation by 0.19%, 0.95% and 0.95% respectively. This shows that the present invention is not only innovative in theory, but also shows significant advantages in practical applications.

[0020] 4. In addition to economic benefits, the present invention also takes into account the need for ecological environmental protection. By accurately controlling the water level of the reservoir, the ecological environmental damage caused by excessive water release or water storage is avoided, which helps to maintain the health and stability of the river ecosystem. This is of great significance for promoting the sustainable use of water resources and achieving a win-win situation for economic and social development and environmental protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0022] Figure 1 A river basin diagram of an embodiment of the present invention;

[0023] Figure 2 The optimized dispatching results of a reservoir from 1959 to 2022 in an embodiment of the present invention;

[0024] Figure 3 This is a comparison of the effects of the rolling simulation strategy in the embodiments of the present invention;

[0025] Figure 4 The following is a comparison of the water level simulation process under various rules in the embodiments of the present invention. DETAILED DESCRIPTION

[0026] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0027] Please refer to Figures 1 to 4 As shown, a reservoir dispatching rule extraction method based on a rolling simulation strategy includes the following steps:

[0028] Step 1: Given the initial water level Z0 of the reservoir during the dispatch period, the dispatch period t = 0;

[0029] Step 2, with initial water level Z t and inflow flow I t As the characteristic factor, the final water level Z t+1 is the target value, training model M t , where M t represents the dispatching rule of the reservoir in the tth dispatching period;

[0030] Step 3: Set the initial water level Z t and inflow flow I t Input model M t , get the decision water level K t+1 ; The decision water level Kt+1 is the final water level predicted by the model Mt, and there is a certain deviation value from the final water level Zt+1 obtained by the optimized scheduling;

[0031] Step 4: Determine the decision water level K t+1 Whether the reservoir dispatching constraints are violated. If so, the constraints are corrected. The specific correction method is to force the decision water level to be equal to the upper limit if it exceeds the upper limit, and to force the decision water level to be equal to the lower limit if it exceeds the lower limit;

[0032] Step 5: Set the decision water level K t+1 Set to the initial water level Z for the next scheduling period t+1 , and let t = t + 1;

[0033] Step 6: Determine whether the scheduling period t meets the end condition. If so, output the final scheduling rule. Otherwise, loop through steps 2 to 6.

[0034] In order to verify the effectiveness of the proposed method, the research objects and data information are proposed:

[0035] Introduction to the research object: A certain river is 441.9 km long, with a drop of 1932 m, an average sand content of 0.437%, a total basin area of ​​26,600 km2, and 13 tributaries. The basin has a typical subtropical plateau monsoon climate, with abundant precipitation, but uneven distribution throughout the year, mainly concentrated in the flood season from May to September, with little interannual variation. The basin has abundant water, rich hydropower resources, concentrated water level drop in the mainstream, good topographic and geological conditions, and has formed a cascade reservoir group with good regulation performance.

[0036] The sub-cascade hydropower stations of Station A, Station B, Station C and Station D on the main stream of a certain river are the key forces for upstream hydropower production. Among the four cascade reservoirs, Station A, Station C and Station D are daily regulation reservoirs that do not have the conditions for extracting long-term dispatching rules, while Station B is a multi-year regulation reservoir that has the conditions for extracting long-term dispatching rules. Therefore, this study mainly focuses on the extraction of dispatching rules for the reservoir at Station B. Figure 3 The spatial distribution of the four cascade reservoirs in a certain river basin is shown. The main characteristics of the reservoir at Station B are shown in Table 1.

[0037] Table 1 Design parameters of Guangzhao Reservoir

[0038]

[0039] Data information introduction: In this study, the time steps for deterministic optimization scheduling and scheduling rule extraction are both monthly. First, based on the long sequence inflow data of the four cascade reservoirs of a certain river from 1959 to 2022, a total of 64 years, the differential evolution algorithm was used to solve the deterministic optimization scheduling results. In the optimization scheduling model, the initial and final water levels of the B Station Reservoir are set to 735m, the output coefficient is set to 9.14, the minimum output is set to 180.2MW, the maximum output is set to the expected maximum output under a specific head, and the maximum discharge is set to the reservoir discharge capacity under a specific head. The optimization scheduling results of the B Station Reservoir from 1959 to 2022 are shown in the figure. Figure 2 shown.

[0040] Afterwards, the optimized scheduling results are used to construct the scheduling rules and extract samples, that is, the optimized scheduling data is divided into two subsets, including a training set and a test set. The training set is a total of 58 years of optimized scheduling data from 1959 to 2016, which is used to train the model parameters; the test set is a 6-year optimized scheduling data from 2017 to 2022, which is used to test the model performance.

[0041] Based on the above objects, eight models for extracting operation rules are proposed, including Long Short-Term Memory Neural Network (LSTM), Artificial Neural Network (ANN), Light Gradient Boosting Machine (LGB), Linear Regression (LR), LSTM-RS, ANN-RS, LGB-RS and LR-RS. Among them, LSTM-RS refers to the LSTM model that adopts the rolling simulation strategy, and ANN-RS, LGB-RS and LR-RS use similar naming methods. In order to obtain better model performance, the Bayesian optimization algorithm is used for hyperparameter optimization of all models. The models used in this study and their related parameters are shown in Table 2. In the scheduling rule extraction experiment, all models use unified input and output, and the input (characteristic factor) is the inflow I t and initial water level Z t , the output (decision variable) is the final water level Z t+1 .

[0042] Table 2 Models and parameters used in the study

[0043]

[0044] Result analysis: This study uses two indicators, power generation and decision water level deviation, to evaluate the pros and cons of the rules extracted by different models. Among them, the decision water level deviation is the root mean square error between the decision water level and the optimal water level. Power generation is used to measure the power generation benefits achieved by the operating rules. The greater the power generation, the better the operating rules, and vice versa. The decision water level deviation indicates the degree of fit between the dispatching rules and the optimal operation process of the reservoir. The smaller the decision water level deviation, the stronger the model's fitting ability, and vice versa. The statistics of power generation of the dispatching rules are shown in Table 3, and the deviation of the decision water level of the dispatching rules is shown in Table 4. In order to present the results more clearly, the best results except the optimization scheme in the table are highlighted in bold.

[0045] Table 3 Statistics of simulated dispatching power generation

[0046]

[0047] Table 4 Statistics of water level deviation in simulated scheduling decision

[0048]

[0049] As shown in Table 3, overall, LSTM-RS has the best annual average power generation performance, which is only 80 million kWh different from the optimization scheme. This result preliminarily confirms the superiority of LSTM-RS proposed in this study. Specifically, compared with ANN-RS, LGB-RS and LR-RS, the annual average power generation of LSTM-RS increased by 0.19%, 0.95% and 0.95%, respectively. Similar to this result, compared with ANN, LGB and LR, the average annual power generation of LSTM increased by 0.27%, 0.88% and 0.84%, respectively. The above results show that the rules extracted based on LSTM have better performance in power generation indicators, indicating that the LSTM model has more outstanding information extraction capabilities and can effectively mine the operating laws in the reservoir optimization scheme. In addition, compared with LSTM, ANN, LGB, LR, the power generation of LSTM-RS, ANN-RS, LGB-RS, and LR-RS models increased by 0.45%, 0.54%, 0.41% and 0.33%, respectively. This result shows that the dispatching rules using the rolling simulation strategy obtain higher power generation benefits, verifying the effectiveness of the rolling simulation strategy in extracting reservoir dispatching rules.

[0050] As shown in Table 4, overall, the decision water level deviation of LSTM-RS is the smallest compared with other models, which is only 0.66m. This result shows that the water level simulation process of LSTM-RS is closest to the ideal solution. Specifically, compared with ANN-RS, LGB-RS and LR-RS, the decision water level deviation of LSTM-RS is reduced by 7.04%, 70.67% and 69.59%, respectively. Similarly, compared with ANN, LGB and LR, the decision water level deviation of LSTM is improved by 18.93%, 43.77% and 39.05%, respectively. The above results show that the rules extracted based on the LSTM model have a small decision water level deviation, which proves the excellent fitting ability of the LSTM-RS model and can accurately capture the water level fluctuations of the optimization scheme. In addition, compared with LSTM, ANN, LGB and LR, the decision water level deviations of LSTM-RS, ANN-RS, LGB-RS and LR-RS are reduced by 60.48%, 65.47%, 24.13% and 20.73%, respectively. This result shows that the rolling simulation strategy can effectively reduce the impact of cumulative decision errors in the dispatching rules, which once again verifies the effectiveness of the proposed strategy in extracting reservoir dispatching rules.

[0051] In order to make a more intuitive comparison, Figure 3 The effect comparison of rolling simulation strategy is shown. Figure 3As shown in the figure, overall, the water level of the model using the rolling simulation strategy is almost consistent with the ideal water level process. Among the models using the rolling simulation strategy, the water level simulation effect of LSTM-RS is the best, and the water level process is closest to the ideal water level. This result is consistent with the results in Table 4, which once again verifies the effectiveness of LSTM-RS in extracting reservoir operation rules. In addition, there is a significant deviation between the water level of the model without the rolling simulation strategy and the ideal water level process, especially from April to June. This is because April to June is a period of declining reservoir water levels, during which the water level fluctuates greatly. These rules are more sensitive to changes in water levels during this period, which amplifies the cumulative impact of decision-making errors and leads to large deviations in decision-making water levels. This result indirectly verifies the superiority of the rolling simulation strategy.

[0052] In order to compare the advantages and disadvantages of the rules extracted by different models, four models, LSTM-RS, ANN-RS, LGB-RS and LR-RS, were selected for further comparison. Figure 4 The water level simulation process of each model in different years is shown. Figure 4 It can be seen that the water level process of LSTM-RS in each year is closest to the optimization scheme. This result once again proves the superiority of the LSTM-RS model in extracting reservoir scheduling rules. The ANN-RS model also has a relatively good water level decision-making process, but its water level process is always slightly lower than that of LSTM-RS, especially in 2019, 2021 and 2022, which affects the power generation efficiency of the reservoir. Unlike the above models, there is a significant deviation between the water level process of LGB-RS and LR-RS and the optimization scheme. The decision water levels of LGB-RS and LR-RS in July, August and September are significantly lower, and the overall fitting accuracy is poor. This also indirectly explains why the indicators of LGB-RS and LR-RS in Tables 3 and 4 perform poorly.

[0053] In summary, the reservoir dispatching rules extracted by the LSTM-RS model have the best comprehensive performance. On the one hand, this result shows that LSTM has a strong fitting ability and can accurately mine the potential laws in the optimization dispatching scheme. On the other hand, it also verifies the practicality of the rolling simulation strategy, which can effectively reduce the accumulation of decision errors in the dispatching rule extraction.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. A reservoir dispatching rule extraction method based on rolling simulation strategy, characterized in that: The following steps are involved: Step 1: Given the initial water level Z0 of the reservoir during the dispatch period, the dispatch period t = 0; Step 2, with the initial water level Z t and inflow flow I t As the characteristic factor, the final water level Z t+1 is the target value, training model M t , where M t represents the dispatching rule of the reservoir in the tth dispatching period; Step 3: Set the initial water level Z t and inflow flow I t Input model M t , get the decision water level K t+1 ; Step 4: Determine the decision water level K t+1 Whether the reservoir operation constraints are violated, and if so, correct them; Step 5: Set the decision water level K t+1 Set to the initial water level Z for the next scheduling period t+1 , and let t = t + 1; Step 6: Determine whether the scheduling period t meets the end condition. If so, output the final scheduling rule. Otherwise, loop through steps 2 to 6.

2. The method for extracting reservoir dispatching rules based on a rolling simulation strategy according to claim 1 is characterized in that: In step 3, the water level K is determined t+1 It is model M t The predicted final water level is different from the final water level Z obtained by optimal scheduling. t+1 There is a certain deviation value.

3. The method for extracting reservoir dispatching rules based on a rolling simulation strategy according to claim 1 is characterized in that: In step 4, the specific method of correction is: if the decision water level exceeds the upper limit value, it is forced to be equal to the upper limit value; if the decision water level exceeds the lower limit value, it is forced to be equal to the lower limit value.

4. The method for extracting reservoir dispatching rules based on a rolling simulation strategy according to claim 1 is characterized in that: In steps 2 and 3, the models used include long short-term memory neural network model, artificial neural network model, light gradient boosting machine model, and linear regression model.

Images