CN106651053A - Dynamic planning improvement algorithm for power generation dispatching of cascade reservoir group - Google Patents

Abstract

The invention discloses an improved dynamic planning algorithm for cascade reservoir group power generation scheduling, and relates to the fields of reservoir operation scheduling and water conservancy information. The method includes: performing post-order traversal on the cascade reservoir group to obtain an optimized sequence; sequentially performing a dynamic programming algorithm on any reservoir A in the optimized sequence to obtain an initial solution sequence for reservoir scheduling of the reservoir A; The above preliminary solution is the discrete quantity of reservoir storage capacity or water level; on the basis of the initial solution sequence, the initial solution sequence is further optimized by using continuous linear programming to obtain the initial solution optimization sequence; the reservoir storage capacity or water level value in the initial solution optimization sequence is used as the described The dispatching decision-making power generation flow of the reservoir A, the outflow runoff of the reservoir A is updated according to the dispatching decision-making power generation flow, and the reservoir dispatching decision of the reservoir A is synthesized based on the reservoir dispatching rules. The invention is suitable for medium and long-term scheduling of cascade reservoir groups, can obtain the actual optimal solution, and takes less time for solving.

Description

An Improved Algorithm for Dynamic Programming of Cascade Reservoir Group Power Generation Scheduling

technical field

The invention relates to the fields of reservoir operation scheduling and water conservancy information, in particular to an improved algorithm for dynamic programming of cascade reservoir group power generation scheduling.

Background technique

Reservoirs are important engineering means for human beings to develop and utilize water resources, and are responsible for many functions and tasks such as flood control, power generation, shipping, and water supply. Reservoir group optimization scheduling is generally to maximize the corresponding single or multiple objectives under the conditions of satisfying various reservoir capacity constraints, power generation output constraints, and river ecological water constraints.

Hydropower is a clean energy, and vigorously developing hydropower is an important measure to achieve sustainable development. Optimizing the joint dispatching of cascade reservoirs to maximize the power generation is a key and difficult issue in reservoir dispatching research at home and abroad. The existing commonly used SLP algorithm and DP algorithm are used to solve the joint dispatching of computational volume reservoir groups. Although the SLP algorithm is easy to converge to the local optimum, the result is not as good as the DP algorithm, but the running time is shorter. The solution result of DP is better than that of SLP, but there is a dimensionality disaster problem, the running time increases geometrically with the increase of the problem scale, and the obtained solution is optimal in a discrete sense, not really optimal.

Contents of the invention

The purpose of the present invention is to provide an improved dynamic programming algorithm for cascade reservoir group power generation dispatching, so as to solve the aforementioned problems in the prior art.

In order to achieve the above object, the improved dynamic programming algorithm of the cascade reservoir group power generation scheduling of the present invention, the method includes:

Step 1, perform post-order traversal on the cascade reservoir group to obtain the optimized sequence;

Step 2, for any reservoir A in the optimization sequence, run the dynamic programming algorithm with the preliminary solution to obtain the initial solution sequence of the reservoir scheduling of the reservoir A; the preliminary solution is the discrete quantity of reservoir storage capacity or water level;

Step 3, on the basis of the initial solution sequence, use continuous linear programming to further optimize the initial solution sequence to obtain the initial solution optimization sequence;

Step 4: Use the reservoir capacity or water level value in the initial solution optimization sequence as the dispatching decision-making power generation flow of the reservoir A, update the outbound runoff of the reservoir A according to the dispatching decision-making power generation flow, and complete the synthesis of reservoir A based on the reservoir dispatching rules Reservoir scheduling decisions.

Preferably, after step 4, it further includes: judging whether the optimization of the last reservoir of the first-level optimization sequence is completed and a scheduling decision is obtained, if yes, ending the optimization; if not, performing step 2 for the next reservoir.

Preferably, when the discrete quantity of water level is selected to run the dynamic programming algorithm, in step 3, on the basis of the initial solution sequence of reservoir A, limit the water level change threshold per period, use continuous linear programming to further optimize the initial solution sequence, and obtain the initial solution optimization Sequence; if reservoir A is a large reservoir, limit the water level change threshold of reservoir A to ±1m per period, and if reservoir A is a small and medium-sized reservoir, limit the water level change threshold of reservoir A to ±0.1m per period.

Preferably, in step 4, the outflow runoff of the reservoir A is updated according to the power generation flow of the scheduling decision, specifically: combining the inflow flow of the reservoir A, the process of the outflow flow of the reservoir A under the initial solution optimization sequence is calculated as the following A reservoir inflow runoff.

Preferably, step 3 is specifically implemented according to the following steps:

S31, according to the water volume balance and the reservoir parameter table, the corresponding water head h, storage capacity V, and power generation flow Q of the reservoir can be obtained when the preliminary solution is dispatched, and at the same time, the change threshold d ₀ is set according to the grid of the running dynamic programming algorithm;

S32, considering that the change of the power generation flow Q _t-1 in the previous time period t-1 will affect the power generation flow Q _t ,...,Q _T in the subsequent time period t,...,T, so let d _t =b(t) d ₀ , where b(t) is the time diffusion coefficient, which can be taken as d _t is the change value of power generation flow in the current period;

S33, take the power generation flow Q ₁ ,...,Q _T as the decision variable, limit the range of the power generation flow Q _t [Q _t -d _t ,Q _t +d _t ], and calculate the period according to the approximately linear B=kQ(t)hΔt Power generation, so with the maximum total power generation as the goal, water balance, non-negative flow, and maximum discharge of the reservoir as constraints, continuous linear programming is performed to obtain the solution of the first iteration, and then the next iteration is calculated;

S34, record Q _k as the solution vector obtained by the kth iteration, judge Whether the threshold is met or the number of iterations exceeds the threshold of the number of iterations. If any judgment condition is met, d ₀ is halved, and then return and execute S32; if any judgment condition is not met, go to S35;

S35, judging Whether it is close to 0, if yes, terminate the solution; if not, return and execute S32.

The beneficial effects of the present invention are:

The method of the present invention takes the maximum power generation in the cascade reservoir group power generation scheduling as the research object, synthesizes the two types of reservoir scheduling algorithms, overcomes the defects of the two types of algorithms, and efficiently finds the optimal decision sequence for the power generation scheduling of the cascade reservoir group , to provide scientific basis and technical support for actual scheduling.

Compared with the prior art, the present invention has the following advantages and beneficial effects:

(1) The prior art generally only adopts dynamic programming (DP) to solve the problem, and the obtained solution is optimal in a discrete sense, not actually optimal. The present invention overcomes this shortcoming and can obtain an actual optimal solution.

(2) Compared with the prior art, the solution time of the present invention is less.

(3) The present invention is suitable for medium and long-term operation of cascade reservoir groups.

Description of drawings

Fig. 1 is a flow diagram of the improved dynamic programming algorithm for cascade reservoir group power generation scheduling;

Figure 2 is a schematic diagram of the relative upstream and downstream relationships of the reservoirs in the Han River Basin.

detailed description

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings. It should be understood that the specific embodiments described here are only used to explain the present invention, and are not intended to limit the present invention.

The technical scheme of the present invention will be further described below through an example of optimization of power generation scheduling of cascade reservoirs in the Han River Basin and in conjunction with the accompanying drawings.

Fig. 1 is the flow chart of this embodiment, and concrete steps are as follows:

Step 1, perform post-order traversal on the cascade reservoir group to obtain the optimized sequence;

Figure 2 shows the relative upstream and downstream relationships of the reservoirs in the Han River Basin. It is easy to know that Cuijiaying is the most downstream reservoir. Therefore, taking Cuijiaying as the root node, the postorder traversal obtains the sequence of calculation of the reservoir group from upstream to downstream: Doulingzi, Eping, Zhoujiayuan , Songshuling, Pankou, Xiaoxuan, Huanglongtan, Danjiangkou, Wangfuzhou, Sanliping, Siping, Cuijiaying.

Step 2, for the current reservoir in the sequence, run the dynamic programming algorithm DP with a larger storage capacity or water level discrete quantity to obtain a preliminary solution; the corresponding operation in the example

The parameters of Doulingzi Reservoir are shown in Table 1, Table 2, Table 3 and Table 4 below; Table 1 is the parameters of Doulingzi Reservoir; Table 2 is the relationship between the water level and storage capacity of Doulingzi Reservoir; Table 3 is the 36 days of Doulingzi Reservoir Inflow data (m ³ /s); Table 4 shows the relationship between the outflow flow and tail water level of Doulingzi Reservoir.

Table 1 Parameters of Doulingzi Reservoir

Table 2 Relationship between water level and storage capacity of Doulingzi Reservoir

Table 3 Inflow data of Doulingzi Reservoir in the past 36 days (m ³ /s)

ten days 1 2 3 4 5 6 7 8 9 flow 5.24 6.24 5.74 5.37 5.16 6.42 5.71 7.55 8.92 ten days 10 11 12 13 14 15 16 17 18 flow 14.5 9.82 7.75 14.5 70.95 11.66 29.71 14.38 10.44 ten days 19 20 twenty one twenty two twenty three twenty four 25 26 27 flow 31.17 164.49 36.44 103.57 29.38 18.16 25.49 139.12 179.97 ten days 28 29 30 31 32 33 34 35 36 flow 35.52 22.57 17.46 14.51 14.26 11.96 9.75 9.57 11.09

Table 4 Relationship between outflow flow and tail water level of Doulingzi Reservoir

The water level of Doulingzi Reservoir is discretized according to a coarse grid of 2m, and the water level regulation sequence S1 is obtained by DP solution. The basic principle of DP solution is:

F(t,j)=max{F(t-1,i)+B(Q(t),i,j)}, F(0,0)=0,

B=kQ(t)h(i,j)Δt

Among them, F(t,j) represents the maximum power generation when the discrete value of current water level is j from 0 to the beginning of period t, i is all discrete water levels in the previous period, F(0,0)=0 is the boundary condition; B is Calculation formula of power generation benefit in t period, Q is power generation flow, h is water head, and k is coefficient. For the specific calculation process, see the program pseudo code:

Step 3, on the basis of the preliminary solution, use continuous linear programming (SLP) for further optimization;

The principle of continuous linear programming (SLP) is: because the water level storage capacity curve is a single-peak convex function, therefore, when the power generation flow Q _i,j changes very little, the water head h _i,j can be considered as a constant, the formula B=kQ( t)h(i,j)Δt is approximately linear. Specific steps are as follows:

S31, according to the water volume balance and the reservoir parameter table, the corresponding water head h, storage capacity V, and power generation flow Q of the reservoir can be obtained when the preliminary solution is dispatched, and at the same time, the change threshold d ₀ is set according to the grid of the running dynamic programming algorithm;

S32, considering that the change of the power generation flow Q _t-1 in the previous time period t-1 will affect the power generation flow Q _t ,...,Q _T in the subsequent time period t,...,T, so let d _t =b(t) d ₀ , where b(t) is the time diffusion coefficient, which can be taken as d _t is the change value of power generation flow in the current period;

S33, take the power generation flow Q ₁ ,...,Q _T as the decision variable, limit the range of the power generation flow Q _t [Q _t -d _t ,Q _t +d _t ], and calculate the period according to the approximately linear B=kQ(t)hΔt Power generation, so with the maximum total power generation as the goal, water balance, non-negative flow, and maximum discharge of the reservoir as constraints, continuous linear programming is performed to obtain the solution of the first iteration, and then the next iteration is calculated;

S34, record Q _k as the solution vector obtained by the kth iteration, judge Whether the threshold is met or the number of iterations exceeds the threshold of the number of iterations. If any judgment condition is met, d ₀ is halved, and then return and execute S32; if any judgment condition is not met, go to S35;

S35, judging Whether it is close to 0, if yes, terminate the solution; if not, return and execute S32.

In this example, on the basis of S1, limit the water level variation threshold to ±1m per period, use continuous linear programming (SLP) to further optimize, and solve the water level scheduling sequence S2; the optimal solution at this time is the optimal scheduling solution for this reservoir.

Step 4: Use the further optimized reservoir capacity or water level value as the reservoir’s scheduling decision-making power generation flow, and update the corresponding outflow runoff

In this example, combined with the inflow flow process, the outflow flow process R under the S2 water level sequence is calculated as the inflow runoff of the next-level reservoir;

Step 5, check whether it is the last reservoir in the sequence, if so, end the optimization, otherwise execute step 2 for the next reservoir in the sequence.

This process ensures that the reservoirs are optimized in sequence from upstream to downstream until the end.

Comparing the reservoir optimization results: the traditional DP algorithm has a total power generation of 1,078,800 kWh of all reservoirs in the cascade when the water level is 1m discrete, while the total power generation of the SLP-DP algorithm is 1,085,100 kWh when the DP water level is 2m discrete, and the program time is The former is 60%, achieving better optimization results in a shorter time.

By adopting the above-mentioned technical scheme disclosed by the present invention, the following beneficial effects are obtained:

The method of the invention combines the algorithm of cascade reservoir group power generation dispatching using continuous linear programming SLP and dynamic programming DP two types of dispatching algorithms. The invention combines the advantages of the two algorithms, and obtains a better solution than the original two algorithms in a shorter time.

The above is only a preferred embodiment of the present invention, and it should be pointed out that for those of ordinary skill in the art, some improvements and modifications can be made without departing from the principle of the present invention. It should be regarded as the protection scope of the present invention.

Claims (5)

1. A dynamic programming improved algorithm of cascade reservoir group power generation dispatching, it is characterized in that, described method comprises: Step 1, perform post-order traversal on the cascade reservoir group to obtain the optimized sequence; Step 2, for any reservoir A in the optimization sequence, run the dynamic programming algorithm with the preliminary solution to obtain the initial solution sequence of the reservoir scheduling of the reservoir A; the preliminary solution is the discrete quantity of reservoir storage capacity or water level; Step 3, on the basis of the initial solution sequence, use continuous linear programming to further optimize the initial solution sequence to obtain the initial solution optimization sequence; Step 4: Use the reservoir storage capacity or water level value in the initial solution optimization sequence as the dispatching decision-making power generation flow of the reservoir A, update the outbound runoff of the reservoir A according to the dispatching decision-making power generation flow, and complete the synthesis of reservoir A based on the reservoir dispatching rules Reservoir scheduling decisions. 2. according to the dynamic programming improved algorithm of the cascade reservoir group power generation dispatching described in claim 1, it is characterized in that, after step 4 also includes: judge whether to finish the optimization of the last reservoir of the first-level optimization sequence and obtain dispatching decision, if yes , end the optimization; if not, perform step 2 for the next reservoir. 3. according to the dynamic programming improvement algorithm of cascade reservoir group power generation dispatching described in claim 1, it is characterized in that, when selecting the discrete quantity of water level to run the dynamic programming algorithm, in step 3, on the basis of the initial solution sequence of reservoir A, limit The threshold of water level changes in each period, using continuous linear programming to further optimize the initial solution sequence, and obtain the initial solution optimization sequence; If reservoir A is a large reservoir, limit the water level change threshold of reservoir A per period to ±1m; if reservoir A is a small and medium-sized reservoir, limit the water level change threshold of reservoir A per period or ±0.1m. 4. according to the dynamic programming improved algorithm of cascade reservoir group power generation dispatching described in claim 1, it is characterized in that, step 4, according to dispatching decision-making generating flow, update the outbound runoff of described reservoir A, specifically: Combined with the inflow flow of reservoir A, calculate the outflow flow process of reservoir A under the initial solution optimization sequence, as the next reservoir inflow runoff. 5. according to the dynamic programming improved algorithm of cascade reservoir group generation dispatching described in claim 1, it is characterized in that, step 3 is specifically realized according to the following steps: S31, according to the water volume balance and the reservoir parameter table, the corresponding water head h, storage capacity V, and power generation flow Q of the reservoir can be obtained when the preliminary solution is dispatched, and at the same time, the change threshold d ₀ is set according to the grid of the running dynamic programming algorithm; S32, considering that the change of the power generation flow Q _t-1 in the previous time period t-1 will affect the power generation flow Q _t ,...,Q _T in the subsequent time period t,...,T, so let d _t =b(t) d ₀ , where b(t) is the time diffusion coefficient, which can be taken as d _t is the change value of power generation flow in the current period; S33, take the power generation flow Q ₁ ,...,Q _T as the decision variable, limit the range of the power generation flow Q _t [Q _t -d _t ,Q _t +d _t ], and calculate the period according to the approximately linear B=kQ(t)hΔt Power generation, so with the maximum total power generation as the goal, water balance, non-negative flow, and maximum discharge of the reservoir as constraints, continuous linear programming is performed to obtain the solution of the first iteration, and then the next iteration is calculated; S34, record Q _k as the solution vector obtained by the kth iteration, judge Whether the threshold is met or the number of iterations exceeds the threshold of the number of iterations. If any judgment condition is met, d ₀ is halved, and then return and execute S32; if any judgment condition is not met, go to S35; S35, judging Whether it is close to 0, if yes, terminate the solution; if not, return and execute S32.