As many as 50,000 voters -- early 20 percent of those registered in Washington--may be unable to vote or forced to cast disputed ballots in the September D.C. primary election because of foul-ups in compiling the city's voter rolls, according to a City Council study released yesterday.

The study said that the city cannot document precisely how many voters are registered in the District. It said that 15,000 names apparently were dropped inexplicably from the rolls between November 1980 and November 1981; that another 30,000 names appear with no indication of party affiliation, which is necessary for voting in the District's closed primaries; and that another 4,000 names apparently appear with the wrong addresses, which could result in voters' showing up at the wrong precincts.

That makes a total of nearly 50,000 voters who are improperly accounted for in city records, according to the study.

Albert J. Beveridge III, chairman of the D. C. Board of Elections and Ethics, said the city will undertake a ward-by-ward mass mailing soon to all persons listed on its computer tapes, at least 280,000 names, to try to correct the problems. But he said he is not sure that the rolls can be corrected by the Sept. 14 primary.

The critical study was prepared for the council's committee on government operations by former D. C. auditor Matthew S. Watson. It was released as the council committee began a two-day hearing into the problems that characterized last November's general election, and how to avoid them in September.

Watson told the committee the elections board "cannot presently document the exact number of active voters" in the District "because there are no internal records to account for new registrations, address or name changes or deletions."

Watson said a November 1980 computer tape showed 288,837 registered voters in the District, but a computer tape prepared for the 1981 elections showed only 273,185 voters, a drop of just over 15,000 voters. Watson and Beveridge agreed that it was extremely unlikely that so many voters had left the city or canceled their registration in one year.

In addition to those apparently dropped, another 30,000 registrations do not designate the voter as a member of any political party or as an independent, according to the committee's preliminary study. This "will result in disqualification of these voters" in the Sept. 14 primary, Watson warned.

Another 4,000 change-of-address requests from voters have not been made, Watson said. Those voters would not be allowed to cast regular ballots in their old precincts.

As the hearing began, Beveridge said it is uncertain whether the city will have time to correct the deficiencies found by the committee. "I'm as confused now as when we started" the inquiry last year, he said.

Recipients of the mailings planned by the board will be asked to check the voter information and correct it if necessary at their nearest libraries. Persons who don't receive the cards will be asked to reregister, Beveridge said.

If the discrepancies and omissions are not corrected through the board's planned mailings, persons who do not appear on the rolls but who believe they are qualified to vote in the District would have to cast disputed votes, called "challenge ballots," in the September primary. These require additional paper work on the part of both the voter and the elections board, and each must be counted and evaluated by hand.

Watson's study said he found 11 different reasons for the election problems, including missing, incorrect or incomplete computer tapes, poor staff work preparations and a failure to periodically audit and analyze the records keep by the board.

Watson's study was the first extensive record of what actually happened in last November's election when many thousands of voters were improperly turned away, left in disgust after delays or were forced to cast cumbersome disputed ballots.

In addition, some of the city's 137 precincts were given incorrect ballots for the hundreds of community Advisory Neighborhood Commission elections that also were being held.

Previous examinations had blamed many of the problems on the city's SHARE computer system, which the elections board uses to prepare some of its data. However, Watson's study indicated that SHARE played only a minor role in compiling much of the erroneous computer data